[10560010] |Morphology (linguistics)
[10560020] |'''Morphology''' is the field of [[linguistics]] that studies the internal structure of words.
[10560030] |(Words as units in the lexicon are the subject matter of [[lexicology]].)
[10560040] |While words are generally accepted as being (with [[clitic]]s) the smallest units of [[syntax]], it is clear that in most (if not all) languages, words can be related to other words by rules.
[10560050] |For example, [[English language|English]] speakers recognize that the words ''dog'', ''dogs'', and ''dog-catcher'' are closely related.
[10560060] |English speakers recognize these relations from their tacit knowledge of the rules of word-formation in English.
[10560070] |They intuit that ''dog'' is to ''dogs'' as ''cat'' is to ''cats''; similarly, ''dog'' is to ''dog-catcher'' as ''dish'' is to ''dishwasher''.
[10560080] |The rules understood by the speaker reflect specific patterns (or regularities) in the way words are formed from smaller units and how those smaller units interact in speech.
[10560090] |In this way, morphology is the branch of linguistics that studies patterns of word-formation within and across languages, and attempts to formulate rules that model the knowledge of the speakers of those languages.
[10560100] |==History ==
[10560110] |The history of morphological analysis dates back to the [[ancient India]]n linguist , who formulated the 3,959 rules of [[Sanskrit]] morphology in the text by using a Constituency Grammar.
[10560120] |The Graeco-Roman grammatical tradition also engaged in morphological analysis.
[10560130] |The term ''morphology'' was coined by [[August Schleicher]] in [[1859]]
[10560140] |== Fundamental concepts ==
[10560150] |=== Lexemes and word forms ===
[10560160] |The distinction between these two senses of "word" is arguably the most important one in morphology.
[10560170] |The first sense of "word," the one in which ''dog'' and ''dogs'' are "the same word," is called '''[[lexeme]]'''.
[10560180] |The second sense is called '''word-form'''.
[10560190] |We thus say that ''dog'' and ''dogs'' are different forms of the same lexeme.
[10560200] |''Dog'' and ''dog-catcher'', on the other hand, are different lexemes; for example, they refer to two different kinds of entities.
[10560210] |The form of a word that is chosen conventionally to represent the canonical form of a word is called a [[lemma (linguistics)|lemma]], or '''citation form'''.
[10560220] |==== Prosodic word vs. morphological word ====
[10560230] |Here are examples from other languages of the failure of a single phonological word to coincide with a single morphological word-form.
[10560240] |In Latin, one way to express the concept of 'NOUN-PHRASE1 and NOUN-PHRASE2' (as in "apples and oranges") is to suffix '-que' to the second noun phrase: "apples oranges-and", as it were.
[10560250] |An extreme level of this theoretical quandary posed by some phonological words is provided by the Kwak'wala language.
[10560260] |In Kwak'wala, as in a great many other languages, meaning relations between nouns, including possession and "semantic case", are formulated by affixes instead of by independent "words".
[10560270] |The three word English phrase, "with his club", where 'with' identifies its dependent noun phrase as an instrument and 'his' denotes a possession relation, would consist of two words or even just one word in many languages.
[10560280] |But affixation for semantic relations in Kwak'wala differs dramatically (from the viewpoint of those whose language is not Kwak'wala) from such affixation in other languages for this reason: the affixes phonologically attach not to the lexeme they pertain to semantically, but to the ''preceding'' lexeme.
[10560290] |Consider the following example (in Kwakw'ala, sentences begin with what corresponds to an English verb):
[10560300] |kwixʔid-i-da bəgwanəmai-χ-a q'asa-s-isi t'alwagwayu
[10560310] |Morpheme by morpheme translation:
[10560320] |kwixʔid-i-da = clubbed-PIVOT-DETERMINER
[10560330] |bəgwanəma-χ-a = man-ACCUSATIVE-DETERMINER
[10560340] |q'asa-s-is = otter-INSTRUMENTAL-3.PERSON.SINGULAR-POSSESSIVE
[10560350] |t'alwagwayu = club.
[10560360] |"the man clubbed the otter with his club"
[10560370] |(Notation notes:
[10560380] |1. accusative case marks an entity that something is done to.
[10560390] |2. determiners are words such as "the", "this", "that".
[10560400] |3. the concept of "pivot" is a theoretical construct that is not relevant to this discussion.)
[10560410] |That is, to the speaker of Kwak'wala, the sentence does not contain the "words" 'him-the-otter' or 'with-his-club' Instead, the markers -''i-da'' (PIVOT-'the'), referring to ''man'', attaches not to ''bəgwanəma'' ('man'), but instead to the "verb"; the markers -''χ-a'' (ACCUSATIVE-'the'), referring to ''otter'', attach to ''bəgwanəma'' instead of to ''q'asa'' ('otter'), etc.
[10560420] |To summarize differently: a speaker of Kwak'wala does ''not'' perceive the sentence to consist of these phonological words:
[10560430] |kwixʔid i-da-bəgwanəma χ-a-q'asa s-isi-t'alwagwayu
[10560440] |"clubbed PIVOT-the-mani hit-the-otter with-hisi-club
[10560450] |A central publication on this topic is the recent volume edited by Dixon and Aikhenvald (2007), examining the mismatch between prosodic-phonological and grammatical definitions of "word" in various Amazonian, Australian Aboriginal, Caucasian, Eskimo, Indo-European, Native North American, and West African languages, and in sign languages.
[10560460] |Apparently, a wide variety of languages make use of the hybrid linguistic unit clitic, possessing the grammatical features of independent words but the prosodic-phonological lack of freedom of bound morphemes.
[10560470] |The intermediate status of clitics poses a considerable challenge to linguistic theory.
[10560480] |=== Inflection vs. word-formation ===
[10560490] |Given the notion of a lexeme, it is possible to distinguish two kinds of morphological rules.
[10560500] |Some morphological rules relate to different forms of the same lexeme; while other rules relate to different lexemes.
[10560510] |Rules of the first kind are called '''[[Inflection|inflectional rules]]''', while those of the second kind are called '''[[word formation|word-formation]]'''.
[10560520] |The English plural, as illustrated by ''dog'' and ''dogs'', is an inflectional rule; compounds like ''dog-catcher'' or ''dishwasher'' provide an example of a word-formation rule.
[10560530] |Informally, word-formation rules form "new words" (that is, new lexemes), while inflection rules yield variant forms of the "same" word (lexeme).
[10560540] |There is a further distinction between two kinds of word-formation: [[Derivation (linguistics)|derivation]] and [[Compound (linguistics)|compounding]].
[10560550] |Compounding is a process of word-formation that involves combining complete word-forms into a single '''compound''' form; ''dog-catcher'' is therefore a compound, because both ''dog'' and ''catcher'' are complete word-forms in their own right before the compounding process has been applied, and are subsequently treated as one form.
[10560560] |Derivation involves [[affix]]ing [[bound morpheme|bound]] (non-independent) forms to existing lexemes, whereby the addition of the affix '''derives''' a new lexeme.
[10560570] |One example of derivation is clear in this case: the word ''independent'' is derived from the word ''dependent'' by prefixing it with the derivational prefix ''in-'', while ''dependent'' itself is derived from the verb ''depend''.
[10560580] |The distinction between inflection and word-formation is not at all clear-cut.
[10560590] |There are many examples where linguists fail to agree whether a given rule is inflection or word-formation.
[10560600] |The next section will attempt to clarify this distinction.
[10560610] |=== Paradigms and morphosyntax ===
[10560620] |A '''paradigm''' is the complete set of related word-forms associated with a given lexeme.
[10560630] |The familiar examples of paradigms are the [[Grammatical conjugation|conjugations]] of verbs, and the [[declension]]s of nouns.
[10560640] |Accordingly, the word-forms of a lexeme may be arranged conveniently into tables, by classifying them according to shared inflectional categories such as [[grammatical tense|tense]], [[grammatical aspect|aspect]], [[grammatical mood|mood]], [[grammatical number|number]], [[grammatical gender|gender]] or [[grammatical case|case]].
[10560650] |For example, the personal pronouns in English can be organized into tables, using the categories of person (1st., 2nd., 3rd.), number (singular vs. plural), gender (masculine, feminine, neuter), and [[grammatical case|case]] (subjective, objective, and possessive).
[10560660] |See [[English personal pronouns]] for the details.
[10560670] |The inflectional categories used to group word-forms into paradigms cannot be chosen arbitrarily; they must be categories that are relevant to stating the [[syntax|syntactic rules]] of the language.
[10560680] |For example, person and number are categories that can be used to define paradigms in English, because English has [[Agreement (linguistics)|grammatical agreement]] rules that require the verb in a sentence to appear in an inflectional form that matches the person and number of the subject.
[10560690] |In other words, the syntactic rules of English care about the difference between ''dog'' and ''dogs'', because the choice between these two forms determines which form of the verb is to be used.
[10560700] |In contrast, however, no syntactic rule of English cares about the difference between ''dog'' and ''dog-catcher'', or ''dependent'' and ''independent''.
[10560710] |The first two are just nouns, and the second two just adjectives, and they generally behave like any other noun or adjective behaves.
[10560720] |An important difference between inflection and word-formation is that inflected word-forms of lexemes are organized into paradigms, which are defined by the requirements of syntactic rules, whereas the rules of word-formation are not restricted by any corresponding requirements of syntax.
[10560730] |Inflection is therefore said to be relevant to syntax, and word-formation is not.
[10560740] |The part of morphology that covers the relationship between [[syntax]] and morphology is called morphosyntax, and it concerns itself with inflection and paradigms, but not with word-formation or compounding.
[10560750] |=== Allomorphy ===
[10560760] |In the exposition above, morphological rules are described as analogies between word-forms: ''dog'' is to ''dogs'' as ''cat'' is to ''cats'', and as ''dish'' is to ''dishes''.
[10560770] |In this case, the analogy applies both to the form of the words and to their meaning: in each pair, the first word means "one of X", while the second "two or more of X", and the difference is always the plural form ''-s'' affixed to the second word, signaling the key distinction between singular and plural entities.
[10560780] |One of the largest sources of complexity in morphology is that this one-to-one correspondence between meaning and form scarcely applies to every case in the language.
[10560790] |In English, we have word form pairs like ''ox/oxen'', ''goose/geese'', and ''sheep/sheep'', where the difference between the singular and the plural is signaled in a way that departs from the regular pattern, or is not signaled at all.
[10560800] |Even cases considered "regular", with the final ''-s'', are not so simple; the ''-s'' in ''dogs'' is not pronounced the same way as the ''-s'' in ''cats'', and in a plural like ''dishes'', an "extra" vowel appears before the ''-s''.
[10560810] |These cases, where the same distinction is effected by alternative forms of a "word", are called '''[[allomorph]]y'''.
[10560820] |Phonological rules constrain which sounds can appear next to each other in a language, and morphological rules, when applied blindly, would often violate phonological rules, by resulting in sound sequences that are prohibited in the language in question.
[10560830] |For example, to form the plural of ''dish'' by simply appending an ''-s'' to the end of the word would result in the form *{{IPA|[dɪʃs]}}, which is not permitted by the [[phonotactics]] of English.
[10560840] |In order to "rescue" the word, a vowel sound is inserted between the root and the plural marker, and {{IPA|[dɪʃəz]}} results.
[10560850] |Similar rules apply to the pronunciation of the ''-s'' in ''dogs'' and ''cats'': it depends on the quality (voiced vs. unvoiced) of the final preceding [[phoneme]].
[10560860] |=== Lexical morphology ===
[10560870] |[[Lexical morphology]] is the branch of morphology that deals with the [[lexicon]], which, morphologically conceived, is the collection of [[lexeme]]s in a language.
[10560880] |As such, it concerns itself primarily with word-formation: derivation and compounding.
[10560890] |== Models of morphology ==
[10560900] |There are three principal approaches to morphology, which each try to capture the distinctions above in different ways.
[10560910] |These are,
[10560920] |* [[Morpheme-based morphology]], which makes use of an [[Item-and-Arrangment (Morphology)|Item-and-Arrangement]] approach.
[10560930] |* [[Lexeme-based morphology]], which normally makes use of an [[Item-and-Process (Morphology)|Item-and-Process]] approach.
[10560940] |* [[Word-based morphology]], which normally makes use of a [[Word-and-paradigm morphology|Word-and-Paradigm]] approach.
[10560950] |Note that while the associations indicated between the concepts in each item in that list is very strong, it is not absolute.
[10560960] |=== Morpheme-based morphology ===
[10560970] |In [[morpheme-based morphology]], word-forms are analyzed as arrangements of [[morpheme]]s.
[10560980] |A '''morpheme''' is defined as the minimal meaningful unit of a language.
[10560990] |In a word like ''independently'', we say that the morphemes are ''in-'', ''depend'', ''-ent'', and ''ly''; ''depend'' is the [[root (linguistics)|root]] and the other morphemes are, in this case, derivational affixes.
[10561000] |In a word like ''dogs'', we say that ''dog'' is the root, and that ''-s'' is an inflectional morpheme.
[10561010] |This way of analyzing word-forms as if they were made of morphemes put after each other like beads on a string, is called [[Item-and-Arrangment (Morphology)|Item-and-Arrangement]].
[10561020] |The morpheme-based approach is the first one that beginners to morphology usually think of, and which laymen tend to find the most obvious.
[10561030] |This is so to such an extent that very often beginners think that morphemes are an inevitable, fundamental notion of morphology, and many five-minute explanations of morphology are, in fact, five-minute explanations of morpheme-based morphology.
[10561040] |This is, however, not so.
[10561050] |The fundamental idea of morphology is that the words of a language are related to each other by different kinds of rules.
[10561060] |Analyzing words as sequences of morphemes is a way of describing these relations, but is not the only way.
[10561070] |In actual academic linguistics, morpheme-based morphology certainly has many adherents, but is by no means the dominant approach.
[10561080] |=== Lexeme-based morphology ===
[10561090] |[[Lexeme-based morphology]] is (usually) an [[Item-and-Process (Morphology)|Item-and-Process]] approach.
[10561100] |Instead of analyzing a word-form as a set of morphemes arranged in sequence, a word-form is said to be the result of applying rules that ''alter'' a word-form or stem in order to produce a new one.
[10561110] |An inflectional rule takes a stem, changes it as is required by the rule, and outputs a word-form; a derivational rule takes a stem, changes it as per its own requirements, and outputs a derived stem; a compounding rule takes word-forms, and similarly outputs a compound stem.
[10561120] |=== Word-based morphology ===
[10561130] |[[Word-based morphology]] is a (usually) [[Word-and-paradigm morphology|Word-and-paradigm]] approach.
[10561140] |This theory takes paradigms as a central notion.
[10561150] |Instead of stating rules to combine morphemes into word-forms, or to generate word-forms from stems, word-based morphology states generalizations that hold between the forms of inflectional paradigms.
[10561160] |The major point behind this approach is that many such generalizations are hard to state with either of the other approaches.
[10561170] |The examples are usually drawn from [[fusional language]]s, where a given "piece" of a word, which a morpheme-based theory would call an inflectional morpheme, corresponds to a combination of grammatical categories, for example, "third person plural."
[10561180] |Morpheme-based theories usually have no problems with this situation, since one just says that a given morpheme has two categories.
[10561190] |Item-and-Process theories, on the other hand, often break down in cases like these, because they all too often assume that there will be two separate rules here, one for third person, and the other for plural, but the distinction between them turns out to be artificial.
[10561200] |Word-and-Paradigm approaches treat these as whole words that are related to each other by [[analogy|analogical]] rules.
[10561210] |Words can be categorized based on the pattern they fit into.
[10561220] |This applies both to existing words and to new ones.
[10561230] |Application of a pattern different than the one that has been used historically can give rise to a new word, such as ''older'' replacing ''elder'' (where ''older'' follows the normal pattern of [[adjective|adjectival]] [[superlative]]s) and ''cows'' replacing ''kine'' (where ''cows'' fits the regular pattern of plural formation).
[10561240] |While a Word-and-Paradigm approach can explain this easily, other approaches have difficulty with phenomena such as this.
[10561250] |== Morphological typology ==
[10561260] |In the 19th century, philologists devised a now classic classification of languages according to their morphology.
[10561270] |According to this typology, some languages are [[isolating language|isolating]], and have little to no morphology; others are [[agglutinating language|agglutinative]], and their words tend to have lots of easily-separable morphemes; while others yet are inflectional or [[fusional language|fusional]], because their inflectional morphemes are said to be "fused" together.
[10561280] |This leads to one bound morpheme conveying multiple pieces of information.
[10561290] |The classic example of an isolating language is [[Chinese language|Chinese]]; the classic example of an agglutinative language is [[Turkish language|Turkish]]; both [[Latin language|Latin]] and [[Greek language|Greek]] are classic examples of fusional languages.
[10561300] |Considering the variability of the world's languages, it becomes clear that this classification is not at all clear-cut, and many languages do not neatly fit any one of these types, and some fit in more than one.
[10561310] |A continuum of complex morphology of language may be adapted when considering languages.
[10561320] |The three models of morphology stem from attempts to analyze languages that more or less match different categories in this typology.
[10561330] |The Item-and-Arrangement approach fits very naturally with agglutinative languages; while the Item-and-Process and Word-and-Paradigm approaches usually address fusional languages.
[10561340] |The reader should also note that the classical typology also mostly applies to inflectional morphology.
[10561350] |There is very little fusion going on with word-formation.
[10561360] |Languages may be classified as synthetic or analytic in their word formation, depending on the preferred way of expressing notions that are not inflectional: either by using word-formation (synthetic), or by using syntactic phrases (analytic).
[10570010] |Named entity recognition
[10570020] |'''Named entity recognition''' (NER) (also known as '''entity identification (EI)''' and '''entity extraction''') is a subtask of [[information extraction]] that seeks to locate and classify atomic elements in text into predefined categories such as the names of persons, organizations, locations, expressions of times, quantities, monetary values, percentages, etc.
[10570030] |For example, a NER system producing [[Message Understanding Conference|MUC]]-style output might [[Metadata|tag]] the sentence,
[10570040] |:''Jim bought 300 shares of Acme Corp. in 2006.''
[10570050] |:''''''''Jim'''''' bought ''''''300'''''' shares of ''''''Acme Corp.'''''' in ''''''2006''''''''.
[10570060] |NER systems have been created that use linguistic [[formal grammar|grammar]]-based techniques as well as [[statistical model]]s.
[10570070] |Hand-crafted grammar-based systems typically obtain better results, but at the cost of months of work by experienced [[Linguistics|linguists]].
[10570080] |Statistical NER systems typically require a large amount of manually [[annotation|annotated]] training data.
[10570090] |Since about 1998, there has been a great deal of interest in entity identification in the [[molecular biology]], [[bioinformatics]], and medical [[natural language processing]] communities.
[10570100] |The most common entity of interest in that domain has been names of genes and gene products.
[10570110] |==Named entity types==
[10570120] |In the expression ''named entity'', the word ''named'' restricts the task to those entities for which one or many [[rigid designator]]s, as defined by [[Saul Kripke|Kripke]], stands for the referent.
[10570130] |For instance, the ''automotive company created by Henry Ford in 1903'' is referred to as ''Ford'' or ''Ford Motor Company''.
[10570140] |Rigid designators include proper names as well as certain natural kind terms like biological species and substances.
[10570150] |There is a general agreement to include [[temporal expressions]] and some numerical expressions such as money and measures in named entities.
[10570160] |While some instances of these types are good examples of rigid designators (e.g., the year 2001) there are also many invalid ones (e.g., I take my vacations in “June”).
[10570170] |In the first case, the year ''2001'' refers to the ''2001st year of the Gregorian calendar''.
[10570180] |In the second case, the month ''June'' may refer to the month of an undefined year (''past June'', ''next June'', ''June 2020'', etc.).
[10570190] |It is arguable that the named entity definition is loosened in such cases for practical reasons.
[10570200] |At least two [[Hierarchy|hierarchies]] of named entity types have been proposed in the literature.
[10570210] |[[BBN Technologies|BBN]] categories [http://www.ldc.upenn.edu/Catalog/docs/LDC2005T33/BBN-Types-Subtypes.html], proposed in 2002, is used for [[Question Answering]] and consists of 29 types and 64 subtypes.
[10570220] |Sekine's extended hierarchy [http://nlp.cs.nyu.edu/ene/], proposed in 2002, is made of 200 subtypes.
[10570230] |==Evaluation==
[10570240] |Benchmarking and evaluations have been performed in the ''[[Message Understanding Conference]]s'' (MUC) organized by [[DARPA]], ''International Conference on Language Resources and Evaluation (LREC)'', ''Computational Natural Language Learning ([[CoNLL]])'' workshops, ''Automatic Content Extraction'' (ACE) organized by [[NIST]], the ''[[Multilingual Entity Task Conference]]'' (MET), ''Information Retrieval and Extraction Exercise'' (IREX) and in ''HAREM'' (Portuguese language only).
[10570250] |[http://aclweb.org/aclwiki/index.php?title=Named_Entity_Recognition_%28State_of_the_art%29 State-of-the-art systems] produce near-human performance.
[10570260] |For instance, the best system entering [http://www.itl.nist.gov/iad/894.02/related_projects/muc/proceedings/muc_7_toc.html MUC-7] scored 93.39% of [[Information_retrieval#F-measure|f-measure]] while human annotators scored 97.60% and 96.95%.
[10580010] |Natural language
[10580020] |In the [[philosophy of language]], a '''natural language''' (or '''ordinary language''') is a [[language]] that is spoken, [[writing|written]], or [[Sign language|signed]] by [[animal]]s for general-purpose communication, as distinguished from [[formal language]]s (such as [[Programming language|computer-programming languages]] or the "languages" used in the study of formal [[logic]], especially [[mathematical logic]]) and from [[constructed language]]s.
[10580030] |== Defining natural language ==
[10580040] |Though the exact definition is debatable, natural language is often contrasted with artificial or [[constructed languages]] such as [[Esperanto]], [[Latino Sexione]], and [[Occidental language|Occidental]].
[10580050] |Linguists have an incomplete understanding of all aspects of the rules underlying natural languages, and these rules are therefore objects of study.
[10580060] |The understanding of natural languages reveals much about not only how language works (in terms of [[syntax]], [[semantics]], [[phonetics]], [[phonology]], etc), but also about how the human [[mind]] and the human [[brain]] process language.
[10580070] |In linguistic terms, 'natural language' only applies to a language that has evolved naturally, and the study of natural language primarily involves native (first language) speakers.
[10580080] |The theory of [[universal grammar]] proposes that all natural languages have certain underlying rules which constrain the structure of the specific grammar for any given language.
[10580090] |While [[grammarians]], writers of dictionaries, and language policy-makers all have a certain influence on the evolution of language, their ability to influence what people think they 'ought' to say is distinct from what people actually say.
[10580100] |Natural language applies to the latter, and is thus a 'descriptive' rather than a 'prescriptive' term.
[10580110] |Thus non-standard language varieties (such as [[African American Vernacular English]]) are considered to be natural while standard language varieties (such as [[Standard American English]]) which are more 'prescripted' can be considered to be at least somewhat artificial or constructed.
[10580120] |== Native language learning ==
[10580130] |The [[learning]] of one's own [[native language]], typically that of one's [[parent]]s, normally occurs spontaneously in early human [[childhood]] and is [[Biology|biologically]] driven.
[10580140] |A crucial role of this process is performed by the [[Nervous system|neural]] activity of a portion of the human [[brain]] known as [[Broca's area]].
[10580150] |There are approximately 7,000 current human languages, and many, if not most seem to share certain properties, leading to the belief in the existence of [[Universal Grammar]], as shown by [[generative grammar]] studies pioneered by the work of [[Noam Chomsky]].
[10580160] |Recently, it has been demonstrated that a dedicated network in the human brain (crucially involving [[Broca's area]], a portion of the left inferior frontal gyrus), is selectively activated by complex verbal structures (but not simple ones) of those languages that meet the Universal Grammar requirements.
[10580170] |== Origins of natural language ==
[10580180] |There is disagreement among anthropologists on when language was first used by humans (or their ancestors).
[10580190] |Estimates range from about two million (2,000,000) years ago, during the time of ''[[Homo habilis]]'', to as recently as forty thousand (40,000) years ago, during the time of [[Cro-Magnon]] man.
[10580200] |However recent evidence suggests modern human language was invented or evolved in Africa prior to the dispersal of humans from Africa around 50,000 years ago.
[10580210] |Since all people including the most isolated indigenous groups such as the [[Andamanese]] or the [[Tasmanian aboriginals]] possess language, then it must have been present in the ancestral populations in Africa before the human population split into various groups to colonize the rest of the world.
[10580220] |Some claim that all nautural languages came out of one single language, known as [[Adamic]].
[10580230] |== Linguistic diversity ==
[10580240] |As of early 2007, there are 6,912 known living human languages.
[10580250] |A "living language" is simply one which is in wide use by a specific group of living people.
[10580260] |The exact number of known living languages will vary from 5,000 to 10,000, depending generally on the precision of one's definition of "language", and in particular on how one classifies [[dialects]].
[10580270] |There are also many dead or [[extinct language]]s.
[10580280] |There is no [[dialect#.22Dialect.22 or .22language.22|clear distinction]] between a language and a [[dialect]], notwithstanding linguist [[Max Weinreich]]'s famous [[aphorism]] that "[[a language is a dialect with an army and navy]]."
[10580290] |In other words, the distinction may hinge on political considerations as much as on cultural differences, distinctive [[writing system]]s, or degree of [[mutual intelligibility]].
[10580300] |It is probably impossible to accurately enumerate the living languages because our worldwide knowledge is incomplete, and it is a "moving target", as explained in greater detail by the [[Ethnologue]]'s Introduction, p. 7 - 8.
[10580310] |With the 15th edition, the 103 newly added languages are not new but reclassified due to refinements in the definition of language.
[10580320] |Although widely considered an [[encyclopedia]], the [[Ethnologue]] actually presents itself as an incomplete catalog, including only named languages that its editors are able to document.
[10580330] |With each edition, the number of catalogued languages has grown.
[10580340] |Beginning with the 14th edition (2000), an attempt was made to include all known living languages.
[10580350] |SIL used an internal 3-letter code fashioned after [[airport code]]s to identify languages.
[10580360] |This was the precursor to the modern [[ISO 639-3]] standard, to which SIL contributed.
[10580370] |The standard allows for over 14,000 languages.
[10580380] |In turn, the 15th edition was revised to conform to the pending ISO 639-3 standard.
[10580390] |Of the catalogued languages, 497 have been flagged as "nearly extinct" due to trends in their usage.
[10580400] |Per the 15th edition, 6,912 living languages are shared by over 5.7 billion speakers. (p. 15)
[10580410] |== Taxonomy ==
[10580420] |The [[Taxonomic classification|classification]] of natural languages can be performed on the basis of different underlying principles (different closeness notions, respecting different properties and relations between languages); important directions of present classifications are:
[10580430] |* paying attention to the historical evolution of languages results in a genetic classification of languages—which is based on genetic relatedness of languages,
[10580440] |* paying attention to the internal structure of languages ([[grammar]]) results in a typological classification of languages—which is based on similarity of one or more components of the language's grammar across languages,
[10580450] |* and respecting geographical closeness and contacts between language-speaking communities results in areal groupings of languages.
[10580460] |The different classifications do not match each other and are not expected to, but the correlation between them is an important point for many [[linguistics|linguistic]] research works.
[10580470] |(There is a parallel to the classification of [[species]] in biological [[phylogenetics]] here: consider [[monophyletic]] vs. [[polyphyletic]] groups of species.)
[10580480] |The task of genetic classification belongs to the field of [[historical-comparative linguistics]], of typological—to [[linguistic typology]].
[10580490] |See also [[Taxonomy]], and [[Taxonomic classification]] for the general idea of classification and taxonomies.
[10580500] |==== Genetic classification ====
[10580510] |The world's languages have been grouped into families of languages that are believed to have common ancestors.
[10580520] |Some of the major families are the [[Indo-European languages]], the [[Afro-Asiatic languages]], the [[Austronesian languages]], and the [[Sino-Tibetan languages]].
[10580530] |The shared features of languages from one family can be due to shared ancestry.
[10580540] |(Compare with [[homology (biology)|homology]] in biology.)
[10580550] |==== Typological classification ====
[10580560] |An example of a typological classification is the classification of languages on the basis of the basic order of the [[verb]], the [[subject (grammar)|subject]] and the [[object (grammar)|object]] in a [[sentence (linguistics)|sentence]] into several types: [[SVO language|SVO]], [[SOV language|SOV]], [[VSO language|VSO]], and so on, languages.
[10580570] |([[English language|English]], for instance, belongs to the [[SVO language]] type.)
[10580580] |The shared features of languages of one type (= from one typological class) may have arisen completely independently.
[10580590] |(Compare with [[analogy (biology)|analogy]] in biology.)
[10580595] |Their cooccurence might be due to the universal laws governing the structure of natural languages—[[language universal]]s.
[10580600] |==== Areal classification ====
[10580610] |The following language groupings can serve as some linguistically significant examples of areal linguistic units, or ''[[sprachbund]]s'': [[Balkan linguistic union]], or the bigger group of [[European languages]]; [[Caucasian languages]]; [[East Asian languages]].
[10580620] |Although the members of each group are not closely [[genetic relatedness of languages|genetically related]], there is a reason for them to share similar features, namely: their speakers have been in contact for a long time within a common community and the languages ''converged'' in the course of the history.
[10580630] |These are called "[[areal feature (linguistics)|areal feature]]s".
[10580640] |One should be careful about the underlying classification principle for groups of languages which have apparently a geographical name: besides areal linguistic units, the [[taxa]] of the genetic classification ([[language family|language families]]) are often given names which themselves or parts of which refer to geographical areas.
[10580650] |== Controlled languages ==
[10580660] |Controlled natural languages are subsets of natural languages whose grammars and dictionaries have been restricted in order to reduce or eliminate both ambiguity and complexity.
[10580670] |The purpose behind the development and implementation of a controlled natural language typically is to aid non-native speakers of a natural language in understanding it, or to ease computer processing of a natural language.
[10580680] |An example of a widely used controlled natural language is [[Simplified English]], which was originally developed for [[aerospace]] industry maintenance manuals.
[10580690] |== Constructed languages and international auxiliary languages ==
[10580700] |Constructed [[international auxiliary language]]s such as [[Esperanto]] and [[Interlingua]] that have [[native speaker]]s are by some also considered natural languages.
[10580710] |However, constructed languages, while they are clearly languages, are not generally considered natural languages.
[10580720] |The problem is that other languages have been used to communicate and evolve in a natural way, while Esperanto has been selectively designed by [[L.L. Zamenhof]] from natural languages, not grown from the natural fluctuations in vocabulary and syntax.
[10580730] |Nor has Esperanto been naturally "standardized" by children's natural tendency to correct for illogical grammar structures in their parents' language, which can be seen in the development of [[pidgin]] languages into [[creole language]]s (as explained by Steven Pinker in [[The Language Instinct]]).
[10580740] |The possible exception to this are true native speakers of such languages.
[10580750] |More substantive basis for this designation is that the vocabulary, grammar, and orthography of Interlingua are natural; they have been standardized and presented by a [[International Auxiliary Language Association|linguistic research body]], but they predated it and are not themselves considered a product of human invention.
[10580760] |Most experts, however, consider Interlingua to be naturalistic rather than natural.
[10580770] |[[Latino Sine Flexione]], a second naturalistic auxiliary language, is also naturalistic in content but is no longer widely spoken.
[10580780] |==Natural Language Processing==
[10580790] |Natural language processing (NLP) is a subfield of artificial intelligence and computational linguistics.
[10580800] |It studies the problems of automated generation and understanding of natural human languages.
[10580810] |Natural-language-generation systems convert information from computer databases into normal-sounding human language.
[10580820] |Natural-language-understanding systems convert samples of human language into more formal representations that are easier for computer programs to manipulate.
[10580830] |== Modalities ==
[10580840] |Natural language manifests itself in modalities other than speech.
[10580850] |=== Sign languages ===
[10580860] |In linguistic terms, sign languages are as rich and complex as any oral language, despite the previously common misconception that they are not "real languages".
[10580870] |Professional linguists have studied many sign languages and found them to have every linguistic component required to be classed as true natural languages.
[10580880] |Sign languages are not [[pantomime]], much as most spoken language is not [[onomatopoeic]].
[10580890] |The signs do tend to exploit iconicity (visual connections with their referents) more than what is common in spoken language, but they are above all conventional and hence generally incomprehensible to non-speakers, just like spoken words and morphemes.
[10580900] |They are not a visual rendition of an oral language either.
[10580910] |They have complex grammars of their own, and can be used to discuss any topic, from the simple and concrete to the lofty and abstract.
[10580920] |=== Written languages ===
[10580930] |In a sense, written language should be distinguished from natural language.
[10580940] |Until recently in the developed world, it was common for many people to be fluent in [[spoken language|spoken]] or [[sign language|signed languages]] and yet remain illiterate; this is still the case in poor countries today.
[10580950] |Furthermore, natural [[language acquisition]] during childhood is largely spontaneous, while [[literacy]] must usually be intentionally acquired.
[10590010] |Natural language processing
[10590020] |'''Natural language processing''' ('''NLP''') is a subfield of [[artificial intelligence]] and [[computational linguistics]].
[10590030] |It studies the problems of automated generation and understanding of [[natural language|natural human languages]].
[10590040] |Natural-language-generation systems convert information from computer databases into normal-sounding human language.
[10590050] |Natural-language-understanding systems convert samples of human language into more formal representations that are easier for [[computer]] programs to manipulate.
[10590060] |==Tasks and limitations==
[10590070] |In theory, natural-language processing is a very attractive method of [[human-computer interaction]].
[10590080] |Early systems such as [[SHRDLU]], working in restricted "[[blocks world]]s" with restricted vocabularies, worked extremely well, leading researchers to excessive optimism, which was soon lost when the systems were extended to more realistic situations with real-world [[ambiguity]] and [[complexity]].
[10590090] |Natural-language understanding is sometimes referred to as an [[AI-complete]] problem, because natural-language recognition seems to require extensive knowledge about the outside world and the ability to manipulate it.
[10590100] |The definition of "[[understanding]]" is one of the major problems in natural-language processing.
[10590110] |==Concrete problems==
[10590120] |Some examples of the problems faced by natural-language-understanding systems:
[10590130] |* The sentences ''We gave the monkeys the bananas because they were hungry'' and ''We gave the monkeys the bananas because they were over-ripe'' have the same surface grammatical structure.
[10590140] |However, the pronoun ''they'' refers to ''monkeys'' in one sentence and ''bananas'' in the other, and it is impossible to tell which without a knowledge of the properties of monkeys and bananas.
[10590150] |* A string of words may be interpreted in different ways.
[10590160] |For example, the string ''Time flies like an arrow'' may be interpreted in a variety of ways:
[10590170] |**The common [[simile]]: ''[[time]]'' moves quickly just like an arrow does;
[10590180] |**measure the speed of flies like you would measure that of an arrow (thus interpreted as an imperative) - i.e. ''(You should) time flies as you would (time) an arrow.'';
[10590190] |**measure the speed of flies like an arrow would - i.e. ''Time flies in the same way that an arrow would (time them).'';
[10590200] |**measure the speed of flies that are like arrows - i.e. ''Time those flies that are like arrows'';
[10590210] |**all of a type of flying insect, "time-flies," collectively enjoys a single arrow (compare ''Fruit flies like a banana'');
[10590220] |**each of a type of flying insect, "time-flies," individually enjoys a different arrow (similar comparison applies);
[10590230] |**A concrete object, for example the magazine, ''[[Time (magazine)|Time]]'', travels through the air in an arrow-like manner.
[10590240] |English is particularly challenging in this regard because it has little [[inflectional morphology]] to distinguish between [[parts of speech]].
[10590250] |* English and several other languages don't specify which word an adjective applies to.
[10590260] |For example, in the string "pretty little girls' school".
[10590270] |** Does the school look little?
[10590280] |** Do the girls look little?
[10590290] |** Do the girls look pretty?
[10590300] |** Does the school look pretty?
[10590310] |* We will often imply additional information in spoken language by the way we place stress on words.
[10590320] |The sentence "I never said she stole my money" demonstrates the importance stress can play in a sentence, and thus the inherent difficulty a natural language processor can have in parsing it.
[10590330] |Depending on which word the speaker places the stress, this sentence could have several distinct meanings:
[10590340] |** "'''I''' never said she stole my money" - Someone else said it, but ''I'' didn't.
[10590350] |** "I '''never''' said she stole my money" - I simply didn't ever say it.
[10590360] |** "I never '''said''' she stole my money" - I might have implied it in some way, but I never explicitly said it.
[10590370] |** "I never said '''she''' stole my money" - I said someone took it; I didn't say it was she.
[10590380] |** "I never said she '''stole''' my money" - I just said she probably borrowed it.
[10590390] |** "I never said she stole '''my''' money" - I said she stole someone else's money.
[10590400] |** "I never said she stole my '''money'''" - I said she stole something, but not my money.
[10590410] |==Subproblems==
[10590420] |; [[Speech segmentation]]:
[10590430] |In most spoken languages, the sounds representing successive letters blend into each other, so the conversion of the analog signal to discrete characters can be a very difficult process.
[10590440] |Also, in [[natural speech]] there are hardly any pauses between successive words; the location of those boundaries usually must take into account [[grammatical]] and [[semantic]] constraints, as well as the [[context]].
[10590450] |; [[Text segmentation]]:
[10590460] |Some written languages like [[Chinese language|Chinese]], [[Japanese language|Japanese]] and [[Thai language|Thai]] do not have single-word boundaries either, so any significant text [[parsing]] usually requires the identification of word boundaries, which is often a non-trivial task.
[10590470] |; [[Word sense disambiguation]]:
[10590480] |Many words have more than one [[meaning]]; we have to select the meaning which makes the most sense in context.
[10590490] |; [[Syntactic ambiguity]]:
[10590500] |The [[grammar]] for [[natural language]]s is [[ambiguous]], i.e. there are often multiple possible [[parse tree]]s for a given sentence.
[10590510] |Choosing the most appropriate one usually requires [[semantics|semantic]] and contextual information.
[10590520] |Specific problem components of syntactic ambiguity include [[sentence boundary disambiguation]].
[10590530] |; Imperfect or irregular input :
[10590540] |Foreign or regional accents and vocal impediments in speech; typing or grammatical errors, [[Optical character recognition|OCR]] errors in texts.
[10590550] |; [[Speech acts]] and plans:
[10590560] |A sentence can often be considered an action by the speaker.
[10590570] |The sentence structure, alone, may not contain enough information to define this action.
[10590580] |For instance, a question is actually the speaker requesting some sort of response from the listener.
[10590590] |The desired response may be verbal, physical, or some combination.
[10590600] |For example, "Can you pass the class?" is a request for a simple yes-or-no answer, while "Can you pass the salt?" is requesting a physical action to be performed.
[10590610] |It is not appropriate to respond with "Yes, I can pass the salt," without the accompanying action (although "No" or "I can't reach the salt" would explain a lack of action).
[10590620] |== Statistical NLP ==
[10590630] |Statistical natural-language processing uses [[stochastic]], [[probabilistic]] and [[statistical]] methods to resolve some of the difficulties discussed above, especially those which arise because longer sentences are highly ambiguous when processed with realistic grammars, yielding thousands or millions of possible analyses.
[10590640] |Methods for disambiguation often involve the use of [[corpus linguistics | corpora]] and [[Markov model]]s.
[10590650] |Statistical NLP comprises all quantitative approaches to automated language processing, including probabilistic modeling, [[information theory]], and [[linear algebra]].
[10590660] |The technology for statistical NLP comes mainly from [[machine learning]] and [[data mining]], both of which are fields of [[artificial intelligence]] that involve learning from data.
[10590670] |==Major tasks in NLP==
[10590680] |* [[Automatic summarization]]
[10590690] |* [[Foreign language reading aid]]
[10590700] |* [[Foreign language writing aid]]
[10590710] |* [[Information extraction]]
[10590720] |* [[Information retrieval]]
[10590730] |* [[Machine translation]]
[10590740] |* [[Named entity recognition]]
[10590750] |* [[Natural language generation]]
[10590760] |* [[Natural language understanding]]
[10590770] |* [[Optical character recognition]]
[10590780] |* [[Question answering]]
[10590790] |* [[Speech recognition]]
[10590800] |* [[Spoken dialogue system]]
[10590810] |* [[Text simplification]]
[10590820] |* [[Text to speech]]
[10590830] |* [[Text-proofing]]
[10590840] |== Evaluation of natural language processing ==
[10590850] |===Objectives===
[10590860] |The goal of NLP evaluation is to measure one or more ''qualities'' of an algorithm or a system, in order to determine if (or to what extent) the system answers the goals of its designers, or the needs of its users.
[10590870] |Research in NLP evaluation has received considerable attention, because the definition of proper evaluation criteria is one way to specify precisely an NLP problem, going thus beyond the vagueness of tasks defined only as ''language understanding'' or ''language generation''.
[10590880] |A precise set of evaluation criteria, which includes mainly evaluation data and evaluation metrics, enables several teams to compare their solutions to a given NLP problem.
[10590890] |===Short history of evaluation in NLP===
[10590900] |The first evaluation campaign on written texts seems to be a campaign dedicated to message understanding in 1987 (Pallet 1998).
[10590910] |Then, the Parseval/GEIG project compared phrase-structure grammars (Black 1991).
[10590920] |A series of campaigns within Tipster project were realized on tasks like summarization, translation and searching (Hirshman 1998).
[10590930] |In 1994, in Germany, the Morpholympics compared German taggers.
[10590940] |Then, the Senseval and Romanseval campaigns were conducted with the objectives of semantic disambiguation.
[10590950] |In 1996, the Sparkle campaign compared syntactic parsers in four different languages (English, French, German and Italian).
[10590960] |In France, the Grace project compared a set of 21 taggers for French in 1997 (Adda 1999).
[10590970] |In 2004, during the [[Technolangue/Easy]] project, 13 parsers for French were compared.
[10590980] |Large-scale evaluation of dependency parsers were performed in the context of the CoNLL shared tasks in 2006 and 2007.
[10590990] |In Italy, the evalita campaign was conducted in 2007 to compare various tools for Italian [http://evalita.itc.it evalita web site].
[10591000] |In France, within the ANR-Passage project (end of 2007), 10 parsers for French were compared [http://atoll.inria.fr/passage/ passage web site].
[10591010] |Adda G., Mariani J., Paroubek P., Rajman M. 1999 L'action GRACE d'évaluation de l'assignation des parties du discours pour le français. Langues vol-2
[10591030] |Black E., Abney S., Flickinger D., Gdaniec C., Grishman R., Harrison P., Hindle D., Ingria R., Jelinek F., Klavans J., Liberman M., Marcus M., Reukos S., Santoni B., Strzalkowski T. 1991 A procedure for quantitatively comparing the syntactic coverage of English grammars. DARPA Speech and Natural Language Workshop
[10591050] |Hirshman L. 1998 Language understanding evaluation: lessons learned from MUC and ATIS. LREC Granada
[10591070] |Pallet D.S. 1998 The NIST role in automatic speech recognition benchmark tests. LREC Granada
[10591090] |===Different types of evaluation===
[10591100] |Depending on the evaluation procedures, a number of distinctions are traditionally made in NLP evaluation.
[10591110] |* Intrinsic vs. extrinsic evaluation
[10591120] |Intrinsic evaluation considers an isolated NLP system and characterizes its performance mainly with respect to a ''gold standard'' result, pre-defined by the evaluators.
[10591130] |Extrinsic evaluation, also called ''evaluation in use'' considers the NLP system in a more complex setting, either as an embedded system or serving a precise function for a human user.
[10591140] |The extrinsic performance of the system is then characterized in terms of its utility with respect to the overall task of the complex system or the human user.
[10591150] |* Black-box vs. glass-box evaluation
[10591160] |Black-box evaluation requires one to run an NLP system on a given data set and to measure a number of parameters related to the quality of the process (speed, reliability, resource consumption) and, most importantly, to the quality of the result (e.g. the accuracy of data annotation or the fidelity of a translation).
[10591170] |Glass-box evaluation looks at the design of the system, the algorithms that are implemented, the linguistic resources it uses (e.g. vocabulary size), etc.
[10591180] |Given the complexity of NLP problems, it is often difficult to predict performance only on the basis of glass-box evaluation, but this type of evaluation is more informative with respect to error analysis or future developments of a system.
[10591190] |* Automatic vs. manual evaluation
[10591200] |In many cases, automatic procedures can be defined to evaluate an NLP system by comparing its output with the gold standard (or desired) one.
[10591210] |Although the cost of producing the gold standard can be quite high, automatic evaluation can be repeated as often as needed without much additional costs (on the same input data).
[10591220] |However, for many NLP problems, the definition of a gold standard is a complex task, and can prove impossible when inter-annotator agreement is insufficient.
[10591230] |Manual evaluation is performed by human judges, which are instructed to estimate the quality of a system, or most often of a sample of its output, based on a number of criteria.
[10591240] |Although, thanks to their linguistic competence, human judges can be considered as the reference for a number of language processing tasks, there is also considerable variation across their ratings.
[10591250] |This is why automatic evaluation is sometimes referred to as ''objective'' evaluation, while the human kind appears to be more ''subjective.''
[10591260] |=== Shared tasks (Campaigns)===
[10591270] |* [[BioCreative]]
[10591280] |* [[Message Understanding Conference]]
[10591290] |* [[Technolangue/Easy]]
[10591300] |* [[Text Retrieval Conference]]
[10591310] |==Standardization in NLP==
[10591320] |An ISO sub-committee is working in order to ease interoperability between [[Lexical resource]]s and NLP programs.
[10591330] |The sub-committee is part of [[ISO/TC37]] and is called ISO/TC37/SC4.
[10591340] |Some ISO standards are already published but most of them are under construction, mainly on lexicon representation (see [[lexical markup framework|LMF]]), annotation and data category registry.
[10600010] |Neural network
[10600020] |Traditionally, the term '''neural network''' had been used to refer to a network or circuit of [[neuron|biological neurons]].
[10600030] |The modern usage of the term often refers to [[artificial neural network]]s, which are composed of [[artificial neuron]]s or nodes.
[10600040] |Thus the term has two distinct usages:
[10600050] |# [[Biological neural network]]s are made up of real biological neurons that are connected or functionally-related in the [[peripheral nervous system]] or the [[central nervous system]].
[10600060] |In the field of [[neuroscience]], they are often identified as groups of neurons that perform a specific physiological function in laboratory analysis.
[10600070] |# [[Artificial neural network]]s are made up of interconnecting artificial neurons (programming constructs that mimic the properties of biological neurons).
[10600080] |Artificial neural networks may either be used to gain an understanding of biological neural networks, or for solving artificial intelligence problems without necessarily creating a model of a real biological system.
[10600090] |This article focuses on the relationship between the two concepts; for detailed coverage of the two different concepts refer to the separate articles: [[Biological neural network]] and [[Artificial neural network]].
[10600100] |==Characterization==
[10600110] |In general a biological neural network is composed of a group or groups of chemically connected or functionally associated neurons.
[10600120] |A single neuron may be connected to many other neurons and the total number of neurons and connections in a network may be extensive.
[10600130] |Connections, called [[synapses]], are usually formed from [[axons]] to [[dendrites]], though dendrodendritic microcircuits and other connections are possible.
[10600140] |Apart from the electrical signaling, there are other forms of signaling that arise from [[neurotransmitter]] diffusion, which have an effect on electrical signaling.
[10600150] |As such, neural networks are extremely complex.
[10600160] |[[Artificial intelligence]] and [[cognitive modeling]] try to simulate some properties of neural networks.
[10600170] |While similar in their techniques, the former has the aim of solving particular tasks, while the latter aims to build mathematical models of biological neural systems.
[10600180] |In the [[artificial intelligence]] field, artificial neural networks have been applied successfully to [[speech recognition]], [[image analysis]] and adaptive [[control]], in order to construct [[software agents]] (in [[Video game|computer and video games]]) or [[autonomous robot]]s.
[10600190] |Most of the currently employed artificial neural networks for artificial intelligence are based on [[statistical estimation]], [[Optimization (mathematics)|optimization]] and [[control theory]].
[10600200] |The [[cognitive modelling]] field involves the physical or mathematical modeling of the behaviour of neural systems; ranging from the individual neural level (e.g. modelling the spike response curves of neurons to a stimulus), through the neural cluster level (e.g. modelling the release and effects of dopamine in the basal ganglia) to the complete organism (e.g. behavioural modelling of the organism's response to stimuli).
[10600210] |==The brain, neural networks and computers==
[10600220] |Neural networks, as used in artificial intelligence, have traditionally been viewed as simplified models of neural processing in the brain, even though the relation between this model and brain biological architecture is debated.
[10600230] |A subject of current research in theoretical neuroscience is the question surrounding the degree of complexity and the properties that individual neural elements should have to reproduce something resembling animal intelligence.
[10600240] |Historically, computers evolved from the [[von Neumann architecture]], which is based on sequential processing and execution of explicit instructions.
[10600250] |On the other hand, the origins of neural networks are based on efforts to model information processing in biological systems, which may rely largely on parallel processing as well as implicit instructions based on recognition of patterns of 'sensory' input from external sources.
[10600260] |In other words, at its very heart a neural network is a complex statistical processor (as opposed to being tasked to sequentially process and execute).
[10600270] |==Neural networks and artificial intelligence==
[10600280] |An ''artificial neural network'' (ANN), also called a ''simulated neural network'' (SNN) or commonly just ''neural network'' (NN) is an interconnected group of [[artificial neuron]]s that uses a [[mathematical model|mathematical or computational model]] for [[information processing]] based on a [[connectionism|connectionistic]] approach to [[computation]].
[10600290] |In most cases an ANN is an [[adaptive system]] that changes its structure based on external or internal information that flows through the network.
[10600300] |In more practical terms neural networks are [[non-linear]] [[statistical]] [[data modeling]] or [[decision making]] tools.
[10600310] |They can be used to model complex relationships between inputs and outputs or to [[Pattern recognition|find patterns]] in data.
[10600320] |===Background===
[10600330] |An [[artificial neural network]] involves a network of simple processing elements ([[artificial neurons]]) which can exhibit complex global behaviour, determined by the connections between the processing elements and element parameters.
[10600340] |One classical type of artificial neural network is the [[Hopfield net]].
[10600350] |In a neural network model simple [[Node (neural networks)|nodes]], which can be called variously "neurons", "neurodes", "Processing Elements" (PE) or "units", are connected together to form a network of nodes — hence the term "neural network".
[10600360] |While a neural network does not have to be adaptive ''per se'', its practical use comes with algorithms designed to alter the strength (weights) of the connections in the network to produce a desired signal flow.
[10600370] |In modern [[Neural network software|software implementations]] of artificial neural networks the approach inspired by biology has more or less been abandoned for a more practical approach based on statistics and signal processing.
[10600380] |In some of these systems neural networks, or parts of neural networks (such as [[artificial neuron]]s) are used as components in larger systems that combine both adaptive and non-adaptive elements.
[10600390] |The concept of a neural network appears to have first been proposed by [[Alan Turing]] in his 1948 paper "Intelligent Machinery".
[10600400] |===Applications===
[10600410] |The utility of artificial neural network models lies in the fact that they can be used to infer a function from observations and also to use it.
[10600420] |This is particularly useful in applications where the complexity of the data or task makes the design of such a function by hand impractical.
[10600430] |====Real life applications====
[10600440] |The tasks to which artificial neural networks are applied tend to fall within the following broad categories:
[10600450] |*[[Function approximation]], or [[regression analysis]], including [[time series prediction]] and modelling.
[10600460] |*[[Statistical classification|Classification]], including [[Pattern recognition|pattern]] and sequence recognition, novelty detection and sequential decision making.
[10600470] |*[[Data processing]], including filtering, clustering, [[blind signal separation]] and compression.
[10600480] |Application areas include system identification and control (vehicle control, process control), game-playing and decision making (backgammon, chess, racing), pattern recognition (radar systems, face identification, object recognition, etc.), sequence recognition (gesture, speech, handwritten text recognition), medical diagnosis, financial applications, [[data mining]] (or knowledge discovery in databases, "KDD"), visualization and [[e-mail spam]] filtering.
[10600490] |===Neural network software===
[10600500] |''Main article:'' [[Neural network software]]
[10600510] |'''Neural network software''' is used to [[Simulation|simulate]], [[research]], [[Software development|develop]] and apply [[artificial neural network]]s, [[biological neural network]]s and in some cases a wider array of [[adaptive system]]s.
[10600520] |====Learning paradigms====
[10600530] |There are three major learning paradigms, each corresponding to a particular abstract learning task.
[10600540] |These are [[supervised learning]], [[unsupervised learning]] and [[reinforcement learning]].
[10600550] |Usually any given type of network architecture can be employed in any of those tasks.
[10600560] |;Supervised learning
[10600570] |In [[supervised learning]], we are given a set of example pairs and the aim is to find a function in the allowed class of functions that matches the examples.
[10600580] |In other words, we wish to ''infer'' how the mapping implied by the data and the cost function is related to the mismatch between our mapping and the data.
[10600590] |;Unsupervised learning
[10600600] |In [[unsupervised learning]] we are given some data , and a cost function which is to be minimized which can be any function of and the network's output, .
[10600610] |The cost function is determined by the task formulation.
[10600620] |Most applications fall within the domain of [[estimation problems]] such as [[statistical modeling]], [[Data compression|compression]], [[Mail filter|filtering]], [[blind source separation]] and [[data clustering|clustering]].
[10600630] |;Reinforcement learning
[10600640] |In [[reinforcement learning]], data is usually not given, but generated by an agent's interactions with the environment.
[10600650] |At each point in time , the agent performs an action and the environment generates an observation and an instantaneous cost , according to some (usually unknown) dynamics.
[10600660] |The aim is to discover a ''policy'' for selecting actions that minimises some measure of a long-term cost, i.e. the expected cumulative cost.
[10600670] |The environment's dynamics and the long-term cost for each policy are usually unknown, but can be estimated.
[10600680] |ANNs are frequently used in reinforcement learning as part of the overall algorithm.
[10600690] |Tasks that fall within the paradigm of reinforcement learning are [[control]] problems, [[game]]s and other [[sequential decision making]] tasks.
[10600700] |====Learning algorithms====
[10600710] |There are many algorithms for training neural networks; most of them can be viewed as a straightforward application of [[Optimization (mathematics)|optimization]] theory and [[statistical estimation]].
[10600720] |[[Evolutionary computation]] methods, [[simulated annealing]], [[Expectation-Maximization|expectation maximization]] and [[non-parametric methods]] are among other commonly used methods for training neural networks.
[10600730] |See also [[machine learning]].
[10600740] |Recent developments in this field also saw the use of [[particle swarm optimization]] and other [[swarm intelligence]] techniques used in the training of neural networks.
[10600750] |==Neural networks and neuroscience==
[10600760] |Theoretical and [[computational neuroscience]] is the field concerned with the theoretical analysis and computational modeling of biological neural systems.
[10600770] |Since neural systems are intimately related to cognitive processes and behaviour, the field is closely related to cognitive and behavioural modeling.
[10600780] |The aim of the field is to create models of biological neural systems in order to understand how biological systems work.
[10600790] |To gain this understanding, neuroscientists strive to make a link between observed biological processes (data), biologically plausible mechanisms for neural processing and learning ([[biological neural network]] models) and theory (statistical learning theory and [[information theory]]).
[10600800] |=== Types of models ===
[10600810] |Many models are used in the field, each defined at a different level of abstraction and trying to model different aspects of neural systems.
[10600820] |They range from models of the short-term behaviour of [[biological neuron models|individual neurons]], through models of how the dynamics of neural circuitry arise from interactions between individual neurons, to models of how behaviour can arise from abstract neural modules that represent complete subsystems.
[10600830] |These include models of the long-term and short-term plasticity of neural systems and its relation to learning and memory, from the individual neuron to the system level.
[10600840] |===Current research===
[10600850] |While initially research had been concerned mostly with the electrical characteristics of neurons, a particularly important part of the investigation in recent years has been the exploration of the role of [[neuromodulators]] such as [[dopamine]], [[acetylcholine]], and [[serotonin]] on behaviour and learning.
[10600860] |[[Biophysics|Biophysical]] models, such as [[BCM theory]], have been important in understanding mechanisms for [[synaptic plasticity]], and have had applications in both computer science and neuroscience.
[10600870] |Research is ongoing in understanding the computational algorithms used in the brain, with some recent biological evidence for [[radial basis networks]] and [[neural backpropagation]] as mechanisms for processing data.
[10600880] |==History of the neural network analogy==
[10600890] |The concept of neural networks started in the late-1800s as an effort to describe how the human mind performed.
[10600900] |These ideas started being applied to computational models with the [[Perceptron]].
[10600910] |In early 1950s [[Friedrich Hayek]] was one of the first to posit the idea of [[spontaneous order]] in the brain arising out of decentralized networks of simple units (neurons).
[10600920] |In the late 1940s, [[Donald Hebb]] made one of the first hypotheses for a mechanism of neural plasticity (i.e. learning), [[Hebbian learning]].
[10600930] |Hebbian learning is considered to be a 'typical' unsupervised learning rule and it (and variants of it) was an early model for [[long term potentiation]].
[10600940] |The [[Perceptron]] is essentially a linear classifier for classifying data specified by parameters and an output function .
[10600950] |Its parameters are adapted with an ad-hoc rule similar to stochastic steepest gradient descent.
[10600960] |Because the [[inner product]] is a [[linear operator]] in the input space, the Perceptron can only perfectly classify a set of data for which different classes are [[linearly separable]] in the input space, while it often fails completely for non-separable data.
[10600970] |While the development of the algorithm initially generated some enthusiasm, partly because of its apparent relation to biological mechanisms, the later discovery of this inadequacy caused such models to be abandoned until the introduction of non-linear models into the field.
[10600980] |The [[Cognitron]] (1975) was an early multilayered neural network with a training algorithm.
[10600990] |The actual structure of the network and the methods used to set the interconnection weights change from one neural strategy to another, each with its advantages and disadvantages.
[10601000] |Networks can propagate information in one direction only, or they can bounce back and forth until self-activation at a node occurs and the network settles on a final state.
[10601010] |The ability for bi-directional flow of inputs between neurons/nodes was produced with the [[Hopfield net|Hopfield's network]] (1982), and specialization of these node layers for specific purposes was introduced through the first [[hybrid neural network|hybrid network]].
[10601020] |The [[connectionism|parallel distributed processing]] of the mid-1980s became popular under the name [[connectionism]].
[10601030] |The rediscovery of the [[backpropagation]] algorithm was probably the main reason behind the repopularisation of neural networks after the publication of "Learning Internal Representations by Error Propagation" in 1986 (Though backpropagation itself dates from 1974).
[10601040] |The original network utilised multiple layers of weight-sum units of the type , where was a [[sigmoid function]] or [[logistic function]] such as used in [[logistic regression]].
[10601050] |Training was done by a form of stochastic steepest gradient descent.
[10601060] |The employment of the chain rule of differentiation in deriving the appropriate parameter updates results in an algorithm that seems to 'backpropagate errors', hence the nomenclature.
[10601070] |However it is essentially a form of gradient descent.
[10601080] |Determining the optimal parameters in a model of this type is not trivial, and steepest gradient descent methods cannot be relied upon to give the solution without a good starting point.
[10601090] |In recent times, networks with the same architecture as the backpropagation network are referred to as [[Multilayer perceptron|Multi-Layer Perceptrons]].
[10601100] |This name does not impose any limitations on the type of algorithm used for learning.
[10601110] |The backpropagation network generated much enthusiasm at the time and there was much controversy about whether such learning could be implemented in the brain or not, partly because a mechanism for reverse signalling was not obvious at the time, but most importantly because there was no plausible source for the 'teaching' or 'target' signal.
[10601120] |==Criticism==
[10601130] |[[A. K. Dewdney]], a former ''[[Scientific American]]'' columnist, wrote in 1997, ''“Although neural nets do solve a few toy problems, their powers of computation are so limited that I am surprised anyone takes them seriously as a general problem-solving tool.”''
[10601140] |(Dewdney, p.82)
[10601150] |Arguments against Dewdney's position are that neural nets have been successfully used to solve many complex and diverse tasks, ranging from autonomously flying aircraft[http://www.nasa.gov/centers/dryden/news/NewsReleases/2003/03-49.html] to detecting credit card fraud[http://www.visa.ca/en/about/visabenefits/innovation.cfm].
[10601160] |Technology writer [[Roger Bridgman]] commented on Dewdney's statements about neural nets:
[10601170] | Neural networks, for instance, are in the dock not only because they have been hyped to high heaven, (what hasn't?) but also because you could create a successful net without understanding how it worked: the bunch of numbers that captures its behaviour would in all probability be "an opaque, unreadable table...valueless as a scientific resource".
[10601180] |In spite of his emphatic declaration that science is not technology, Dewdney seems here to pillory neural nets as bad science when most of those devising them are just trying to be good engineers.
[10601190] |An unreadable table that a useful machine could read would still be well worth having.
[10610010] |N-gram
[10610020] |An '''''n''-gram''' is a sub-sequence of ''n'' items from a given [[sequence]].
[10610025] |''n''-grams are used in various areas of statistical [[natural language processing]] and genetic sequence analysis.
[10610030] |The items in question can be letters, words or [[base pairs]] according to the application.
[10610040] |An ''n''-gram of size 1 is a "[[unigram]]"; size 2 is a "[[bigram]]" (or, more etymologically sound but less commonly used, a "digram"); size 3 is a "[[trigram]]"; and size 4 or more is simply called an "''n''-gram".
[10610050] |Some [[language model]]s built from n-grams are "(''n'' − 1)-order [[Markov_chain|Markov model]]s".
[10610060] |==Examples==
[10610070] |Here are examples of '''''word''''' level 3-grams and 4-grams (and counts of the number of times they appeared) from the [[N-gram#Google_use_of_N-gram|Google n-gram corpus]].
[10610080] |*ceramics collectables collectibles (55)
[10610090] |*ceramics collectables fine (130)
[10610100] |*ceramics collected by (52)
[10610110] |*ceramics collectible pottery (50)
[10610120] |*ceramics collectibles cooking (45)
[10610130] |4-grams
[10610140] |*serve as the incoming (92)
[10610150] |*serve as the incubator (99)
[10610160] |*serve as the independent (794)
[10610170] |*serve as the index (223)
[10610180] |*serve as the indication (72)
[10610190] |*serve as the indicator (120)
[10610200] |==''n''-gram models==
[10610210] |An '''''n''-gram model''' models sequences, notably natural languages, using the statistical properties of ''n''-grams.
[10610220] |This idea can be traced to an experiment by [[Claude Shannon]]'s work in [[information theory]].
[10610230] |His question was, given a sequence of letters (for example, the sequence "for ex"), what is the [[likelihood]] of the next letter?
[10610240] |From training data, one can derive a [[probability distribution]] for the next letter given a history of size : ''a'' = 0.4, ''b'' = 0.00001, ''c'' = 0, ....; where the probabilities of all possible "next-letters" sum to 1.0.
[10610250] |More concisely, an ''n''-gram model predicts based on .
[10610260] |In Probability terms, this is nothing but .
[10610270] |When used for [[language model|language modeling]] independence assumptions are made so that each word depends only on the last ''n'' words.
[10610280] |This [[Markov model]] is used as an approximation of the true underlying language.
[10610290] |This assumption is important because it massively simplifies the problem of learning the language model from data.
[10610300] |In addition, because of the open nature of language, it is common to group words unknown to the language model together.
[10610310] |''n''-gram models are widely used in statistical [[natural language processing]].
[10610320] |In [[speech recognition]], [[phonemes]] and sequences of phonemes are modeled using a ''n''-gram distribution.
[10610330] |For parsing, words are modeled such that each ''n''-gram is composed of ''n'' words.
[10610340] |For [[language recognition]], sequences of letters are modeled for different languages.
[10610350] |For a sequence of words, (for example "the dog smelled like a skunk"), the trigrams would be: "the dog smelled", "dog smelled like", "smelled like a", and "like a skunk".
[10610360] |For sequences of characters, the 3-grams (sometimes referred to as "trigrams") that can be generated from "good morning" are "goo", "ood", "od ", "d m", " mo", "mor" and so forth.
[10610370] |Some practitioners preprocess strings to remove spaces, most simply collapse whitespace to a single space while preserving paragraph marks.
[10610380] |Punctuation is also commonly reduced or removed by preprocessing.
[10610385] |''n''-grams can also be used for sequences of words or, in fact, for almost any type of data.
[10610390] |They have been used for example for extracting features for clustering large sets of satellite earth images and for determining what part of the Earth a particular image came from.
[10610400] |They have also been very successful as the first pass in genetic sequence search and in the identification of which species short sequences of DNA were taken from.
[10610410] |N-gram models are often criticized because they lack any explicit representation of long range dependency.
[10610420] |While it is true that the only explicit dependency range is (n-1) tokens for an n-gram model, it is also true that the effective range of dependency is significantly longer than this although long range correlations drop exponentially with distance for any Markov model.
[10610430] |Alternative Markov language models that incorporate some degree of local state can exhibit very long range dependencies.
[10610440] |This is often done using hand-crafted state variables that represent, for instance, the position in a sentence, the general topic of discourse or a grammatical state variable.
[10610450] |Some of the best parsers of English currently in existence are roughly of this form.
[10610460] |Another criticism that has been leveled is that Markov models of language, including n-gram models, do not explicitly capture the performance/competence distinction introduced by [[Noam Chomsky]].
[10610470] |This criticism fails to explain why parsers that are the best at parsing text seem to uniformly lack any such distinction and most even lack any clear distinction between semantics and syntax.
[10610480] |Most proponents of n-gram and related language models opt for a fairly pragmatic approach to language modeling that emphasizes empirical results over theoretical purity.
[10610490] |==''n''-grams for approximate matching==
[10610500] |''n''-grams can also be used for efficient approximate matching.
[10610510] |By converting a sequence of items to a set of ''n''-grams, it can be embedded in a [[vector space]] (in other words, represented as a [[histogram]]), thus allowing the sequence to be compared to other sequences in an efficient manner.
[10610520] |For example, if we convert strings with only letters in the English alphabet into 3-grams, we get a -dimensional space (the first dimension measures the number of occurrences of "aaa", the second "aab", and so forth for all possible combinations of three letters).
[10610530] |Using this representation, we lose information about the string.
[10610540] |For example, both the strings "abcba" and "bcbab" give rise to exactly the same 2-grams.
[10610550] |However, we know empirically that if two strings of real text have a similar vector representation (as measured by [[dot product|cosine distance]]) then they are likely to be similar.
[10610560] |Other metrics have also been applied to vectors of ''n''-grams with varying, sometimes better, results.
[10610570] |For example [[z-score]]s have been used to compare documents by examining how many standard deviations each ''n''-gram differs from its mean occurrence in a large collection, or [[text corpus]], of documents (which form the "background" vector).
[10610580] |In the event of small counts, the [[g-score]] may give better results for comparing alternative models.
[10610590] |It is also possible to take a more principled approach to the statistics of ''n''-grams, modeling similarity as the likelihood that two strings came from the same source directly in terms of a problem in [[Bayesian inference]].
[10610600] |==Other applications==
[10610610] |''n''-grams find use in several areas of computer science, [[computational linguistics]], and applied mathematics.
[10610620] |They have been used to:
[10610630] |* design [[kernel (mathematics)|kernels]] that allow [[machine learning]] algorithms such as [[support vector machine]]s to learn from string data
[10610640] |* find likely candidates for the correct spelling of a misspelled word
[10610650] |* improve compression in [[data compression|compression algorithms]] where a small area of data requires ''n''-grams of greater length
[10610660] |* assess the probability of a given word sequence appearing in text of a language of interest in pattern recognition systems, [[speech recognition]], OCR ([[optical character recognition]]), [[Intelligent Character Recognition]] ([[ICR]]), [[machine translation]] and similar applications
[10610670] |* improve retrieval in [[information retrieval]] systems when it is hoped to find similar "documents" (a term for which the conventional meaning is sometimes stretched, depending on the data set) given a single query document and a database of reference documents
[10610680] |* improve retrieval performance in genetic sequence analysis as in the [[BLAST]] family of programs
[10610690] |* identify the language a text is in or the species a small sequence of DNA was taken from
[10610700] |* predict letters or words at random in order to create text, as in the [[dissociated press]] algorithm.
[10610710] |== Bias-versus-variance trade-off ==
[10610720] |What goes into picking the ''n'' for the ''n''-gram?
[10610730] |There are problems of balance weight between ''infrequent grams'' (for example, if a proper name appeared in the training data) and ''frequent grams''.
[10610740] |Also, items not seen in the training data will be given a [[probability]] of 0.0 without [[smoothing]].
[10610750] |For unseen but plausible data from a sample, one can introduce [[pseudocount]]s.
[10610760] |Pseudocounts are generally motivated on Bayesian grounds.
[10610770] |=== Smoothing techniques ===
[10610780] |* [[Linear interpolation]] (e.g., taking the [[weighted mean]] of the unigram, bigram, and trigram)
[10610790] |* [[Good-Turing]] discounting
[10610800] |* [[Witten-Bell discounting]]
[10610810] |* [[Katz's back-off model]] (trigram)
[10610820] |==Google use of N-gram==
[10610830] |[[Google]] uses n-gram models for a variety of R&D projects, such as [[statistical machine translation]], [[speech recognition]], [[Spell checker|checking spelling]], [[entity detection]], and [[information extraction|data mining]].
[10610840] |In September of 2006 [http://googleresearch.blogspot.com/2006/08/all-our-n-gram-are-belong-to-you.html Google announced] that they made their n-grams [http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2006T13 public] at the [[Linguistic Data Consortium]] ([http://www.ldc.upenn.edu/ LDC]).
[10620010] |Noun
[10620020] |In [[linguistics]], a '''noun''' is a member of a large, [[open class (linguistics)|open]] [[lexical category]] whose members can occur as the main word in the [[subject (grammar)|subject]] of a [[clause]], the [[object (grammar)|object]] of a [[verb]], or the object of a [[preposition]].
[10620030] |Lexical categories are defined in terms of how their members combine with other kinds of expressions.
[10620040] |The syntactic rules for nouns differ from language to language.
[10620050] |In [[English language|English]], nouns may be defined as those words which can occur with articles and [[adjective|attributive adjectives]] and can function as the [[phrase|head]] of a [[noun phrase]].
[10620060] |In [[traditional grammar|traditional]] English grammar, the noun is one of the eight [[parts of speech]].
[10620070] |==History==
[10620080] |The word comes from the [[Latin]] ''nomen'' meaning "[[name]]".
[10620090] |Word classes like nouns were first described by the Sanskrit grammarian [[Panini (grammarian)|{{IAST|Pāṇini}}]] and ancient Greeks like [[Dionysios Thrax]]; and were defined in terms of their [[morphology (linguistics)|morphological]] properties.
[10620100] |For example, in Ancient Greek, nouns inflect for [[case (grammar)|grammatical case]], such as dative or accusative.
[10620110] |[[Verb]]s, on the other hand, inflect for [[grammatical tense|tenses]], such as past, present or future, while nouns do not.
[10620120] |[[Aristotle]] also had a notion of ''onomata'' (nouns) and ''rhemata'' (verbs) which, however, does not exactly correspond with modern notions of nouns and verbs.
[10620130] |Vinokurova 2005 has a more detailed discussion of the historical origin of the notion of a noun.
[10620140] |==Different definitions of nouns==
[10620150] |Expressions of [[natural language]] have properties at different levels.
[10620160] |They have ''formal'' properties, like what kinds of [[morphology (linguistics)|morphological]] [[prefix]]es or [[suffix]]es they take and what kinds of other expressions they combine with; but they also have [[semantics|semantic]] properties, i.e. properties pertaining to their meaning.
[10620170] |The definition of a noun at the outset of this page is thus a ''formal'', traditional grammatical definition.
[10620180] |That definition, for the most part, is considered uncontroversial and furnishes the propensity for certain language users to effectively distinguish most nouns from non-nouns.
[10620190] |However, it has the disadvantage that it does not apply to nouns in all languages.
[10620200] |For example in [[Russian language|Russian]], there are no definite articles, so one cannot define nouns as words that are modified by definite articles.
[10620210] |There are also several attempts of defining nouns in terms of their [[semantics|semantic]] properties.
[10620220] |Many of these are controversial, but some are discussed below.
[10620230] |===Names for things===
[10620240] |In [[Traditional grammar|traditional school grammars]], one often encounters the definition of nouns that they are all and only those expressions that refer to a ''person'', ''place'', ''thing'', ''event'', ''substance'', ''quality'', or ''idea'', etc.
[10620250] |This is a ''semantic'' definition.
[10620260] |It has been criticized by contemporary linguists as being uninformative.
[10620270] |Contemporary linguists generally agree that one cannot successfully define nouns (or other grammatical categories) in terms of what sort of ''object in the world'' they ''[[reference|refer]] to'' or ''[[signification|signify]]''.
[10620280] |Part of the [[conundrum]] is that the definition makes use of relatively ''general'' nouns ("thing", "phenomenon", "event") to define what nouns ''are''.
[10620290] |The existence of such ''general'' nouns demonstrates that nouns refer to entities that are organized in [[taxonomy|taxonomic]] [[hierarchies]].
[10620300] |But other kinds of expressions are also organized into such structured taxonomic relationships.
[10620310] |For example the verbs "stroll","saunter", "stride", and "tread" are more specific words than the more ''general'' "walk".
[10620320] |Moreover, "walk" is more specific than the verb "move", which, in turn, is less general than "change".
[10620330] |But it is unlikely that such taxonomic relationships can be used to ''define'' nouns and verbs.
[10620340] |We cannot ''define'' verbs as those words that refer to "changes" or "states", for example, because the nouns ''change'' and ''state'' probably refer to such things, but, of course, aren't verbs.
[10620350] |Similarly, nouns like "invasion", "meeting", or "collapse" refer to things that are "done" or "happen".
[10620360] |In fact, an influential [[theory]] has it that verbs like "kill" or "die" refer to events, which is among the sort of thing that nouns are supposed to refer to.
[10620370] |The point being made here is not that this view of verbs is wrong, but rather that this property of verbs is a poor basis for a ''definition'' of this category, just like the property of ''having wheels'' is a poor basis for a definition of cars (some things that have wheels, such as my suitcase or a jumbo jet, aren't cars).
[10620380] |Similarly, adjectives like "yellow" or "difficult" might be thought to refer to qualities, and adverbs like "outside" or "upstairs" seem to refer to places, which are also among the sorts of things nouns can refer to.
[10620390] |But verbs, adjectives and adverbs are not nouns, and nouns aren't verbs, adjectives or adverbs.
[10620400] |One might argue that "definitions" of this sort really rely on speakers' prior intuitive knowledge of what nouns, verbs and adjectives are, and, so don't really add anything over and beyond this.
[10620410] |Speakers' intuitive knowledge of such things might plausibly be based on ''formal'' criteria, such as the traditional grammatical definition of English nouns aforementioned.
[10620420] |===Prototypically referential expressions===
[10620430] |Another semantic definition of nouns is that they are ''prototypically referential.''
[10620440] |That definition is also not very helpful in distinguishing actual nouns from verbs.
[10620450] |But it may still correctly identify a core property of nounhood.
[10620460] |For example, we will tend to use nouns like "fool" and "car" when we wish to refer to fools and cars, respectively.
[10620470] |The notion that this is '''prototypical''' reflects the fact that such nouns can be used, even though nothing with the corresponding property is referred to:
[10620480] |:John is no '''fool'''.
[10620490] |:If I had a '''car''', I'd go to Marrakech.
[10620500] |The first sentence above doesn't refer to any fools, nor does the second one refer to any particular car.
[10620510] |===Predicates with identity criteria===
[10620520] |The British logician [[Peter Thomas Geach]] proposed a very subtle semantic definition of nouns.
[10620530] |He noticed that adjectives like "same" can modify nouns, but no other kinds of parts of speech, like [[verbs]] or [[adjectives]].
[10620540] |Not only that, but there also doesn't seem to be any ''other'' expressions with similar meaning that can modify verbs and adjectives.
[10620550] |Consider the following examples.
[10620560] |: Good: John and Bill participated in the '''same''' fight.
[10620570] |: Bad:
[10620580] |*John and Bill '''samely''' fought.
[10620590] |There is no English adverb "samely".
[10620600] |In some other languages, like Czech, however there are adverbs corresponding to "samely".
[10620610] |Hence, in Czech, the translation of the last sentence would be fine; however, it would mean that John and Bill fought ''in the same way'': not that they participated in the ''same fight''.
[10620620] |Geach proposed that we could explain this, if nouns denote logical [[predicate (grammar)|predicate]]s with '''identity criteria'''.
[10620630] |An identity criterion would allow us to conclude, for example, that "person x at time 1 is ''the same person'' as person y at time 2".
[10620640] |Different nouns can have different identity criteria.
[10620650] |A well known example of this is due to Gupta:
[10620660] |:National Airlines transported 2 million '''passengers''' in 1979.
[10620670] |:National Airlines transported (at least) 2 million '''persons''' in 1979.
[10620680] |Given that, in general, all passengers are persons, the last sentence above ought to follow logically from the first one.
[10620690] |But it doesn't.
[10620700] |It is easy to imagine, for example, that on average, every person who travelled with National Airlines in 1979, travelled with them twice.
[10620710] |In that case, one would say that the airline transported 2 million ''passengers'' but only 1 million ''persons''.
[10620720] |Thus, the way that we count ''passengers'' isn't necessarily the same as the way that we count ''persons''.
[10620730] |Put somewhat differently: At two different times, ''you'' may correspond to two distinct ''passengers'', even though you are one and the same person.
[10620740] |For a precise definition of ''identity criteria'', see Gupta.
[10620750] |Recently, Baker has proposed that Geach's definition of nouns in terms of identity criteria allows us to ''explain'' the characteristic properties of nouns.
[10620760] |He argues that nouns can co-occur with (in-)definite articles and numerals, and are "prototypically referential" ''because'' they are all and only those [[parts of speech]] that provide identity criteria.
[10620770] |Baker's proposals are quite new, and linguists are still evaluating them.
[10620780] |==Classification of nouns in English==
[10620790] |===Proper nouns and common nouns===
[10620800] |''Proper nouns'' (also called ''[[proper name]]s'') are nouns representing unique entities (such as ''London'', ''Universe'' or ''John''), as distinguished from common nouns which describe a class of entities (such as ''city'', ''planet'' or ''person'').
[10620810] |In [[English language|English]] and most other languages that use the [[Latin alphabet]], proper nouns are usually [[capitalization|capitalized]].
[10620820] |Languages differ in whether most elements of multiword proper nouns are capitalised (e.g., American English ''House of Representatives'') or only the initial element (e.g., Slovenian ''Državni zbor'' 'National Assembly').
[10620830] |In [[German language|German]], nouns of all types are capitalized.
[10620840] |The convention of capitalizing ''all'' nouns was previously used in English, but ended circa 1800.
[10620850] |In America, the shift in capitalization is recorded in several noteworthy documents.
[10620860] |The end (but not the beginning) of the [[United States Declaration of Independence#Annotated text of the Declaration|Declaration of Independence]] (1776) and all of the [[United States Constitution|Constitution]] (1787) show nearly all nouns capitalized, the [[United States Bill of Rights#Text of the Bill of Rights|Bill of Rights]] (1789) capitalizes a few common nouns but not most of them, and the [[Thirteenth Amendment to the United States Constitution|Thirteenth Constitutional Amendment]] (1865) only capitalizes proper nouns.
[10620870] |Sometimes the same word can function as both a common noun and a proper noun, where one such entity is special.
[10620880] |For example the common noun ''god'' denotes all deities, while the proper noun ''God'' references the [[monotheism|monotheistic]] [[God]] specifically.
[10620890] |Owing to the essentially arbitrary nature of [[Orthography|orthographic]] classification and the existence of variant authorities and adopted [[Style guide|''house styles'']], questionable capitalization of words is not uncommon, even in respected newspapers and magazines.
[10620900] |Most publishers, however, properly require ''consistency'', at least within the same document, in applying their specified standard.
[10620910] |The common meaning of the word or words constituting a proper noun may be unrelated to the object to which the proper noun refers.
[10620920] |For example, someone might be named "Tiger Smith" despite being neither a [[tiger]] nor a [[smith (metalwork)|smith]].
[10620930] |For this reason, proper nouns are usually not [[translation|translated]] between languages, although they may be [[transliteration|transliterated]].
[10620940] |For example, the German surname ''Knödel'' becomes ''Knodel'' or ''Knoedel'' in English (not the literal ''Dumpling'').
[10620950] |However, the [[Transliteration|transcription]] of place names and the names of [[monarch]]s, [[pope]]s, and non-contemporary [[author]]s is common and sometimes universal.
[10620960] |For instance, the [[Portuguese language|Portuguese]] word ''Lisboa'' becomes ''[[Lisbon]]'' in [[English language|English]]; the English ''London'' becomes ''Londres'' in French; and the [[ancient Greek|Greek]] ''Aristotelēs'' becomes [[Aristotle]] in English.
[10620970] |===Countable and uncountable nouns===
[10620980] |''Count nouns'' are common nouns that can take a [[plural]], can combine with [[numerals]] or [[quantifiers]] (e.g. "one", "two", "several", "every", "most"), and can take an indefinite article ("a" or "an").
[10620990] |Examples of count nouns are "chair", "nose", and "occasion".
[10621000] |''Mass nouns'' (or ''non-count nouns'') differ from count nouns in precisely that respect: they can't take plural or combine with number words or quantifiers.
[10621010] |Examples from English include "laughter", "cutlery", "helium", and "furniture".
[10621020] |For example, it is not possible to refer to "a furniture" or "three furnitures".
[10621030] |This is true even though the pieces of furniture comprising "furniture" could be counted.
[10621040] |Thus the distinction between mass and count nouns shouldn't be made in terms of what sorts of things the nouns ''refer'' to, but rather in terms of how the nouns ''present'' these entities.
[10621050] |===Collective nouns===
[10621060] |''Collective nouns'' are nouns that refer to ''groups'' consisting of more than one individual or entity, even when they are inflected for the [[Grammatical number|singular]].
[10621070] |Examples include "committee", "herd", and "school" (of herring).
[10621080] |These nouns have slightly different grammatical properties than other nouns.
[10621090] |For example, the [[noun phrases]] that they [[head (syntax)|head]] can serve as the [[subject (grammar)|subject]] of a [[collective predicate]], even when they are inflected for the singular.
[10621100] |A [[collective predicate]] is a predicate that normally can't take a singular subject.
[10621110] |An example of the latter is "talked to each other".
[10621120] |:Good: The '''boys''' talked to each other.
[10621130] |:Bad: *The '''boy''' talked to each other.
[10621140] |:Good: The '''committee''' talked to each other.
[10621150] |===Concrete nouns and abstract nouns===
[10621160] |''Concrete nouns'' refer to [[physical bodies]] which you use at least one of your [[sense]]s to observe.
[10621170] |For instance, "chair", "apple", or "Janet".
[10621180] |''Abstract nouns'' on the other hand refer to [[abstract object]]s, that is ideas or concepts, such as "justice" or "hate".
[10621190] |While this distinction is sometimes useful, the boundary between the two of them is not always clear; consider, for example, the noun "art".
[10621200] |In English, many abstract nouns are formed by adding noun-forming suffixes ("-ness", "-ity", "-tion") to adjectives or verbs.
[10621210] |Examples are "happiness", "circulation" and "serenity".
[10621220] |==Nouns and pronouns==
[10621230] |[[Noun phrase]]s can typically be replaced by [[pronoun]]s, such as "he", "it", "which", and "those", in order to avoid repetition or explicit identification, or for other reasons.
[10621240] |For example, in the sentence "Janet thought that he was weird", the word "he" is a pronoun standing in place of the name of the person in question.
[10621250] |The English word ''one'' can replace parts of [[noun phrase]]s, and it sometimes stands in for a noun.
[10621260] |An example is given below:
[10621270] |: John's car is newer than ''the one'' that Bill has.
[10621280] |But ''one'' can also stand in for bigger subparts of a noun phrase.
[10621290] |For example, in the following example, ''one'' can stand in for ''new car''.
[10621300] |: This new car is cheaper than ''that one''.
[10621310] |==Substantive as a word for "noun"==
[10621320] |Starting with old [[Latin language|Latin]] grammars, many European languages use some form of the word ''substantive'' as the basic term for noun.
[10621330] |Nouns in the dictionaries of such languages are demarked by the abbreviation "s" instead of "n", which may be used for proper nouns instead.
[10621340] |This corresponds to those grammars in which nouns and adjectives phase into each other in more areas than, for example, the English term [[Predicative_adjective#Predicative_adjective|predicate adjective]] entails.
[10621350] |In French and Spanish, for example, adjectives frequently act as nouns referring to people who have the characteristics of the adjective.
[10621360] |An example in English is:
[10621370] |: The ''poor'' you have always with you.
[10621380] |Similarly, an adjective can also be used for a whole group or organization of people:
[10621390] |: The Socialist ''International''.
[10621400] |Hence, these words are substantives that are usually adjectives in English.
[10630010] |Ontology (information science)
[10630020] |In both [[computer science]] and [[information science]], an '''ontology''' is a formal representation of a set of concepts within a [[Domain of discourse|domain]] and the relationships between those concepts.
[10630030] |It is used to [[Reasoning|reason]] about the properties of that domain, and may be used to define the domain.
[10630040] |Ontologies are used in [[artificial intelligence]], the [[Semantic Web]], [[software engineering]], [[biomedical informatics]], [[library science]], and [[information architecture]] as a form of [[knowledge representation]] about the world or some part of it.
[10630050] |Common components of ontologies include:
[10630060] |* Individuals: instances or objects (the basic or "ground level" objects)
[10630070] |* [[Class]]es: [[set (computer science)|set]]s, collections, concepts or types of objects
[10630080] |* [[Attribute (computing)|Attribute]]s: properties, features, characteristics, or parameters that objects (and classes) can have
[10630090] |* [[Relation (mathematics)|Relations]]: ways that classes and objects can be related to one another
[10630100] |* Function terms: complex structures formed from certain relations that can be used in place of an individual term in a statement
[10630110] |* Restrictions: formally stated descriptions of what must be true in order for some assertion to be accepted as input
[10630120] |* Rules: statements in the form of an if-then (antecedent-consequent) sentence that describe the logical inferences that can be drawn from an assertion in a particular form
[10630130] |* Axioms: assertions (including rules) in a logical form that together comprise the overall theory that the ontology describes in its domain of application.
[10630140] |This definition differs from that of "axioms" in generative grammar and formal logic.
[10630150] |In these disciplines, axioms include only statements asserted as ''a priori'' knowledge.
[10630160] |As used here, "axioms" also include the theory derived from axiomatic statements.
[10630170] |* [[Event (philosophy)|Events]]: the changing of attributes or relations
[10630180] |Ontologies are commonly encoded using [[ontology language]]s.
[10630190] |== Elements ==
[10630200] |Contemporary ontologies share many structural similarities, regardless of the language in which they are expressed.
[10630210] |As mentioned above, most ontologies describe individuals (instances), classes (concepts), attributes, and relations.
[10630220] |In this section each of these components is discussed in turn.
[10630230] |=== Individuals ===
[10630240] |Individuals (instances) are the basic, "ground level" components of an ontology.
[10630250] |The individuals in an ontology may include concrete objects such as people, animals, tables, automobiles, molecules, and planets, as well as abstract individuals such as numbers and words.
[10630260] |Strictly speaking, an ontology need not include any individuals, but one of the general purposes of an ontology is to provide a means of classifying individuals, even if those individuals are not explicitly part of the ontology.
[10630270] |In formal extensional ontologies, only the utterances of words and numbers are considered individuals – the numbers and names themselves are classes.
[10630280] |In a 4D ontology, an individual is identified by its spatio-temporal extent.
[10630290] |Examples of formal extensional ontologies are [[ISO 15926]] and the model in development by the [[IDEAS Group]].
[10630300] |=== Classes ===
[10630310] |Classes – concepts that are also called ''type'', ''sort'', ''category'', and ''kind'' – are abstract groups, sets, or collections of objects.
[10630320] |They may contain individuals, other classes, or a combination of both.
[10630330] |Some examples of classes:
[10630340] |* ''Person'', the class of all people
[10630350] |* ''Vehicle'', the class of all vehicles
[10630360] |* ''Car'', the class of all cars
[10630370] |* ''Class'', representing the class of all classes
[10630380] |* ''Thing'', representing the class of all things
[10630390] |Ontologies vary on whether classes can contain other classes, whether a class can belong to itself, whether there is a universal class (that is, a class containing everything), etc.
[10630400] |Sometimes restrictions along these lines are made in order to avoid certain well-known [[paradox]]es.
[10630410] |The classes of an ontology may be [[extensional]] or [[intensional]] in nature.
[10630420] |A class is extensional if and only if it is characterized solely by its membership.
[10630430] |More precisely, a class C is extensional if and only if for any class C', if C' has exactly the same members as C, then C and C' are identical.
[10630440] |If a class does not satisfy this condition, then it is intensional.
[10630450] |While extensional classes are more well-behaved and well-understood mathematically, as well as less problematic philosophically, they do not permit the fine grained distinctions that ontologies often need to make.
[10630460] |For example, an ontology may want to distinguish between the class of all creatures with a kidney and the class of all creatures with a heart, even if these classes happen to have exactly the same members.
[10630470] |In the upper ontologies mentioned above, the classes are defined intensionally.
[10630480] |Intensionally defined classes usually have necessary conditions associated with membership in each class.
[10630490] |Some classes may also have sufficient conditions, and in those cases the combination of necessary and sufficient conditions make that class a fully ''defined'' class.
[10630500] |Importantly, a class can subsume or be subsumed by other classes; a class subsumed by another is called a ''subclass'' of the subsuming class.
[10630510] |For example, ''Vehicle'' subsumes ''Car'', since (necessarily) anything that is a member of the latter class is a member of the former.
[10630520] |The subsumption relation is used to create a hierarchy of classes, typically with a maximally general class like ''Thing'' at the top, and very specific classes like ''2002 Ford Explorer'' at the bottom.
[10630530] |The critically important consequence of the subsumption relation is the inheritance of properties from the parent (subsuming) class to the child (subsumed) class.
[10630540] |Thus, anything that is necessarily true of a parent class is also necessarily true of all of its subsumed child classes.
[10630550] |In some ontologies, a class is only allowed to have one parent (''single inheritance''), but in most ontologies, classes are allowed to have any number of parents (''multiple inheritance''), and in the latter case all necessary properties of each parent are inherited by the subsumed child class.
[10630560] |Thus a particular class of animal (''HouseCat'') may be a child of the class ''Cat'' and also a child of the class ''Pet''.
[10630570] |A partition is a set of related classes and associated rules that allow objects to be placed into the appropriate class.
[10630580] |For example, to the right is the partial diagram of an ontology that has a partition of the ''Car'' class into the classes ''2-Wheel Drive'' and ''4-Wheel Drive''.
[10630590] |The partition rule determines if a particular car is placed in the ''2-Wheel Drive'' or the ''4-Wheel Drive'' class.
[10630600] |If the partition rule(s) guarantee that a single ''Car'' cannot be in both classes, then the partition is called a disjoint partition.
[10630610] |If the partition rules ensure that every concrete object in the super-class is an instance of at least one of the partition classes, then the partition is called an exhaustive partition.
[10630620] |=== Attributes ===
[10630630] |Objects in the ontology can be described by assigning attributes to them.
[10630640] |Each attribute has at least a name and a value, and is used to store information that is specific to the object it is attached to.
[10630650] |For example the Ford Explorer object has attributes such as:
[10630660] |* ''Name'': Ford Explorer
[10630670] |* ''Number-of-doors'': 4
[10630680] |* ''Engine'': {4.0L, 4.6L}
[10630690] |* ''Transmission'': 6-speed
[10630700] |The value of an attribute can be a complex [[data type]]; in this example, the value of the attribute called ''Engine'' is a list of values, not just a single value.
[10630710] |If you did not define attributes for the concepts you would have either a [[taxonomy]] (if [[hyponym]] relationships exist between concepts) or a '''controlled vocabulary'''.
[10630720] |These are useful, but are not considered true ontologies.
[10630730] |===Relationships===
[10630740] |An important use of attributes is to describe the relationships (also known as relations) between objects in the ontology.
[10630750] |Typically a relation is an attribute whose value is another object in the ontology.
[10630760] |For example in the ontology that contains the Ford Explorer and the [[Ford Bronco]], the Ford Bronco object might have the following attribute:
[10630770] |* ''Successor'': Ford Explorer
[10630780] |This tells us that the Explorer is the model that replaced the Bronco.
[10630790] |Much of the power of ontologies comes from the ability to describe these relations.
[10630800] |Together, the set of relations describes the [[semantics]] of the domain.
[10630810] |The most important type of relation is the [[subsumption]] relation (''is-[[superclass]]-of'', the converse of ''[[is-a]]'', ''is-subtype-of'' or ''is-[[subclass]]-of'').
[10630820] |This defines which objects are members of classes of objects.
[10630830] |For example we have already seen that the Ford Explorer ''is-a'' 4-wheel drive, which in turn ''is-a'' Car:
[10630840] |The addition of the is-a relationships has created a hierarchical [[taxonomy]]; a tree-like structure (or, more generally, a [[partially ordered set]]) that clearly depicts how objects relate to one another.
[10630850] |In such a structure, each object is the 'child' of a 'parent class' (Some languages restrict the is-a relationship to one parent for all nodes, but many do not).
[10630860] |Another common type of relations is the [[meronymy]] relation, written as ''part-of'', that represents how objects combine together to form composite objects.
[10630870] |For example, if we extended our example ontology to include objects like Steering Wheel, we would say that "Steering Wheel is-part-of Ford Explorer" since a steering wheel is one of the components of a Ford Explorer.
[10630880] |If we introduce meronymy relationships to our ontology, we find that this simple and elegant tree structure quickly becomes complex and significantly more difficult to interpret manually.
[10630890] |It is not difficult to understand why; an entity that is described as 'part of' another entity might also be 'part of' a third entity.
[10630900] |Consequently, entities may have more than one parent.
[10630910] |The structure that emerges is known as a [[directed acyclic graph]] (DAG).
[10630920] |As well as the standard is-a and part-of relations, ontologies often include additional types of relation that further refine the semantics they model.
[10630930] |These relations are often domain-specific and are used to answer particular types of question.
[10630940] |For example in the domain of automobiles, we might define a ''made-in'' relationship which tells us where each car is built.
[10630950] |So the Ford Explorer is ''made-in'' [[Louisville, Kentucky|Louisville]].
[10630960] |The ontology may also know that Louisville is-in [[Kentucky]] and Kentucky is-a state of the [[United States|USA]].
[10630970] |Software using this ontology could now answer a question like "which cars are made in the U.S.?"
[10630980] |== Domain ontologies and upper ontologies ==
[10630990] |A domain ontology (or domain-specific ontology) models a specific domain, or part of the world.
[10631000] |It represents the particular meanings of terms as they apply to that domain.
[10631010] |For example the word ''[[card]]'' has many different meanings.
[10631020] |An ontology about the domain of [[poker]] would model the "[[playing card]]" meaning of the word, while an ontology about the domain of [[computer hardware]] would model the "[[punch card]]" and "[[video card]]" meanings.
[10631030] |An [[Upper ontology (computer science)|upper ontology]] (or foundation ontology) is a model of the common objects that are generally applicable across a wide range of domain ontologies.
[10631040] |It contains a [[core glossary]] in whose terms objects in a set of domains can be described.
[10631050] |There are several standardized upper ontologies available for use, including [[Dublin Core]], [[General Formal Ontology|GFO]], [[OpenCyc]]/[[ResearchCyc]], [[Suggested Upper Merged Ontology|SUMO]], and [http://www.loa-cnr.it/DOLCE.html DOLCE]l.
[10631060] |[[WordNet]], while considered an upper ontology by some, is not an ontology: it is a unique combination of a [[taxonomy]] and a controlled vocabulary (see above, under Attributes).
[10631070] |The [[Gellish]] ontology is an example of a combination of an upper and a domain ontology.
[10631080] |Since domain ontologies represent concepts in very specific and often eclectic ways, they are often incompatible.
[10631090] |As systems that rely on domain ontologies expand, they often need to merge domain ontologies into a more general representation.
[10631100] |This presents a challenge to the ontology designer.
[10631110] |Different ontologies in the same domain can also arise due to different perceptions of the domain based on cultural background, education, ideology, or because a different representation language was chosen.
[10631120] |At present, merging ontologies is a largely manual process and therefore time-consuming and expensive.
[10631130] |Using a foundation ontology to provide a common definition of core terms can make this process manageable.
[10631140] |There are studies on generalized techniques for merging ontologies, but this area of research is still largely theoretical.
[10631150] |== Ontology languages ==
[10631160] |An [[ontology language]] is a [[formal language]] used to encode the ontology.
[10631170] |There are a number of such languages for ontologies, both proprietary and standards-based:
[10631180] |* [[Web Ontology Language|OWL]] is a language for making ontological statements, developed as a follow-on from [[Resource Description Framework|RDF]] and [[RDFS]], as well as earlier ontology language projects including [[Ontology Inference Layer|OIL]], [[DARPA Agent Markup Language|DAML]] and [[DAMLplusOIL|DAML+OIL]].
[10631190] |OWL is intended to be used over the [[World Wide Web]], and all its elements (classes, properties and individuals) are defined as RDF [[resource (Web)|resources]], and identified by [[Uniform Resource Identifier|URI]]s.
[10631200] |* [[KIF]] is a syntax for [[first-order logic]] that is based on [[S-expression]]s.
[10631210] |* The [[Cyc]] project has its own ontology language called [[CycL]], based on [[first-order predicate calculus]] with some higher-order extensions.
[10631220] |* [[Rule Interchange Format]] (RIF) and [[F-Logic]] combine ontologies and rules.
[10631230] |* The [[Gellish]] language includes rules for its own extension and thus integrates an ontology with an ontology language.
[10631240] |== Relation to the philosophical term ==
[10631250] |The term ''ontology'' has its origin in [[ontology|philosophy]], where it is the name of one fundamental branch of [[metaphysics]], concerned with analyzing various types or modes of ''existence'', often with special attention to the relations between particulars and universals, between intrinsic and extrinsic properties, and between essence and existence.
[10631260] |According to [[Tom Gruber]] at [[Stanford University]], the meaning of ''ontology'' in the context of computer science is “a description of the concepts and relationships that can exist for an [[Software agent|agent]] or a community of agents.”
[10631270] |He goes on to specify that an ontology is generally written, “as a set of definitions of formal vocabulary.”
[10631280] |What ontology has in common in both computer science and philosophy is the representation of entities, ideas, and events, along with their properties and relations, according to a system of categories.
[10631290] |In both fields, one finds considerable work on problems of ontological relativity (e.g. [[Quine]] and [[Kripke]] in philosophy, [[John F. Sowa|Sowa]] and [[Nicola Guarino|Guarino]] in computer science (Top-level ontological categories.
[10631310] |By: Sowa, John F.
[10631320] |In International Journal of Human-Computer Studies, v. 43 (November/December 1995) p. 669-85.), and debates concerning whether a normative ontology is viable (e.g. debates over [[foundationalism]] in philosophy, debates over the [[Cyc]] project in AI).
[10631330] |Differences between the two are largely matters of focus.
[10631340] |Philosophers are less concerned with establishing fixed, controlled vocabularies than are researchers in computer science, while computer scientists are less involved in discussions of first principles (such as debating whether there are such things as fixed essences, or whether entities must be ontologically more primary than processes).
[10631350] |During the second half of the 20th century, philosophers extensively debated the possible methods or approaches to building ontologies, without actually ''building'' any very elaborate ontologies themselves.
[10631360] |By contrast, computer scientists were building some large and robust ontologies (such as [[WordNet]] and [[Cyc]]) with comparatively little debate over ''how'' they were built.
[10631370] |In the early years of the 21st century, the interdisciplinary project of [[cognitive science]] has been bringing the two circles of scholars closer together.
[10631380] |For example, there is talk of a "computational turn in philosophy" which includes philosophers analyzing the formal ontologies of computer science (sometimes even working directly with the software), while researchers in computer science have been making more references to those philosophers who work on ontology (sometimes with direct consequences for their methods).
[10631390] |Still, many scholars in both fields are uninvolved in this trend of cognitive science, and continue to work independently of one another, pursuing separately their different concerns.
[10631400] |==Resources==
[10631410] |===Examples of published ontologies ===
[10631420] |* [[Dublin Core]], a simple ontology for documents and publishing.
[10631430] |* [[Cyc]] for formal representation of the universe of discourse.
[10631440] |* [[Suggested Upper Merged Ontology]], which is a formal upper ontology
[10631450] |* [http://www.ifomis.org/bfo/ Basic Formal Ontology (BFO)], a formal upper ontology designed to support scientific research
[10631460] |* [[Gellish English dictionary]], an ontology that includes a dictionary and taxonomy that includes an upper ontology and a lower ontology that focusses on industrial and business applications in engineering, technology and procurement.
[10631470] |* [http://www.fb10.uni-bremen.de/anglistik/langpro/webspace/jb/gum/index.htm Generalized Upper Model], a linguistically-motivated ontology for mediating between clients systems and natural language technology
[10631480] |* [[WordNet]] Lexical reference system
[10631490] |* [[OBO Foundry]]: a suite of interoperable reference ontologies in biomedicine.
[10631500] |* The [[Ontology for Biomedical Investigations]] is an open access, integrated ontology for the description of biological and clinical investigations.
[10631510] |* [http://colab.cim3.net/file/work/SICoP/ontac/COSMO/ COSMO]: An OWL ontology that is a merger of the basic elements of the OpenCyc and SUMO ontologies, with additional elements.
[10631520] |* [[Gene Ontology]] for [[genomics]]
[10631530] |* [http://pir.georgetown.edu/pro/ PRO], the Protein Ontology of the Protein Information Resource, Georgetown University.
[10631540] |* [http://proteinontology.info/ Protein Ontology] for [[proteomics]]
[10631550] |* [http://sig.biostr.washington.edu/projects/fm/AboutFM.html Foundational Model of Anatomy] for human anatomy
[10631560] |* [[SBO]], the Systems Biology Ontology, for computational models in biology
[10631570] |* [http://www.plantontology.org/ Plant Ontology] for plant structures and growth/development stages, etc.
[10631580] |* [[CIDOC|CIDOC CRM]] (Conceptual Reference Model) - an ontology for "[[cultural heritage]] information".
[10631590] |* [http://www.linguistics-ontology.org/gold.html GOLD ] ('''G'''eneral '''O'''ntology for [[descriptive linguistics|'''L'''inguistic '''D'''escription ]])
[10631600] |* [http://www.landcglobal.com/pages/linkbase.php Linkbase] A formal representation of the biomedical domain, founded upon [http://www.ifomis.org/bfo/ Basic Formal Ontology (BFO)].
[10631610] |* [http://www.loa-cnr.it/Ontologies.html Foundational, Core and Linguistic Ontologies]
[10631620] |* [[ThoughtTreasure]] ontology
[10631630] |* [[LPL]] Lawson Pattern Language
[10631640] |* [[TIME-ITEM]] Topics for Indexing Medical Education
[10631650] |* [[POPE]] Purdue Ontology for Pharmaceutical Engineering
[10631660] |* [[IDEAS Group]] A formal ontology for enterprise architecture being developed by the Australian, Canadian, UK and U.S. Defence Depts. [http://www.ideasgroup.org The IDEAS Group Website]
[10631670] |* [http://www.eden-study.org/articles/2007/problems-ontology-programs_ao.pdf program abstraction taxonomy]
[10631680] |* [http://sweet.jpl.nasa.gov/ SWEET] Semantic Web for Earth and Environmental Terminology
[10631690] |* [http://www.cellcycleontology.org/ CCO] The Cell-Cycle Ontology is an application ontology that represents the cell cycle
[10631700] |===Ontology libraries===
[10631710] |The development of ontologies for the Web has led to the apparition of services providing lists or directories of ontologies with search facility.
[10631720] |Such directories have been called ontology libraries.
[10631730] |The following are static libraries of human-selected ontologies.
[10631740] |* The [http://www.daml.org/ontologies/ DAML Ontology Library] maintains a legacy of ontologies in DAML.
[10631750] |* The [http://protegewiki.stanford.edu/index.php/Protege_Ontology_Library Protege Ontology Library] contains a set of owl, Frame-based and other format ontologies.
[10631760] |* [http://www.schemaweb.info/ SchemaWeb] is a directory of RDF schemata expressed in RDFS, OWL and DAML+OIL.
[10631770] |The following are both directories and search engines.
[10631780] |They include crawlers searching the Web for well-formed ontologies.
[10631790] |* [[Swoogle]] is a directory and search engine for all RDF resources available on the Web, including ontologies.
[10631800] |* The [http://olp.dfki.de/OntoSelect/ OntoSelect] Ontology Library offers similar services for RDF/S, DAML and OWL ontologies.
[10631810] |* [http://www.w3.org/2004/ontaria/ Ontaria] is a "searchable and browsable directory of semantic web data", with a focus on RDF vocabularies with OWL ontologies.
[10631820] |* The [http://www.obofoundry.org/ OBO Foundry / Bioportal]is a suite of interoperable reference ontologies in biology and biomedicine.