From WordNet, to EuroWordNet, to the Global Wordnet Grid: anchoring languages to universal meaning Piek Vossen VU University Amsterdam 1 What kind of resource is wordnet? • Mostly used database in language technology • Enormous impact in language technology development • Large • Free and downloadable • English 2 WordNet http://wordnet.princeton.edu/ • Developed by George Miller and his team at Princeton University, as the implementation of a mental model of the lexicon • Organized around the notion of a synset: a set of synonyms in a language that represent a single concept • Semantic relations between concepts • Covers over 117,000 concepts and over 150,000 English words Relational model of meaning animal kitten animal man boy man woman cat dog cat meisje boy girl kitten puppy dog puppy woman 4 Wordnet: a network of semantically related words {conveyance;transport} {vehicle} {motor vehicle; automotive vehicle} {car mirror} {armrest} {car door} {doorlock} {car; auto; automobile; machine; motorcar} {bumper} {car window} {cruiser; squad car; patrol car; police car; prowl car} {cab; taxi; hack; taxicab} {hinge; flexible joint} Wordnet Semantic Relations WN 1.5 starting point The ‘synset’ as a weak notion of synonymy: “two expressions are synonymous in a linguistic context C if the substitution of one for the other in C does not alter the truth value.” (Miller et al. 1993) Relations between synsets: Relation POS-combination ANTONYMY adjective-to-adjective verb-to-verb HYPONYMY noun-to-noun verb-to-verb MERONYMY noun-to-noun ENTAILMENT verb-to-verb CAUSE verb-to-verb Example good/bad open/ close car/ vehicle walk/ move head/ nose buy/ pay kill/ die 6 Wordnet Data Model Relations type-of type-of part-of Concepts rec: 12345 1 - financial institute rec: 54321 2 - side of a river rec: 9876 - small string instrument rec: 65438 - musician playing violin rec:42654 - musician rec:35576 1 - string of instrument rec:29551 2 - underwear rec:25876 - string instrument Vocabulary of a language bank 1 2 fiddle violin fiddler violist string 7 Some observations on Wordnet • synsets are more compact representations for concepts than word meanings in traditional lexicons • synonyms and hypernyms are substitutional variants: – begin – commence – I once had a canary. The bird got sick. The poor animal died. • hyponymy and meronymy chains are important transitive relations for predicting properties and explaining textual properties: object -> artifact -> vehicle -> 4-wheeled vehicle -> car • strict separation of part of speech although concepts are closely related (bed – sleep) and are similar (dead – death) • lexicalization patterns reveal important mental structures 8 Lexicalization patterns entity object garbage threat artifact building bird 25 unique beginners organism animal plant waste tree flower basic level church canary dog crocodile rose concepts • balance of two principles: abbey common • predict most features canary • apply to most subclasses • where most concepts are created • amalgamate most parts • most abstract level to draw a pictures 9 Wordnet top level 10 Meronymy & pictures beak tail leg 11 Meronymy & pictures 12 Co-reference constraint in wordnet: Cats cannot be a kind of cats • • • • • • • • • S: (n) cat, true cat (feline mammal usually having thick soft fur and no ability to roar: domestic cats; wildcats) S: (n) guy, cat, hombre, bozo (an informal term for a youth or man) "a nice guy"; "the guy's only doing it for some doll" S: (n) cat (a spiteful woman gossip) "what a cat she is!" S: (n) kat, khat, qat, quat, cat, Arabian tea, African tea (the leaves of the shrub Catha edulis which are chewed like tobacco or used to make tea; has the effect of a euphoric stimulant) "in Yemen kat is used daily by 85% of adults" S: (n) cat-o'-nine-tails, cat (a whip with nine knotted cords) "British sailors feared the cat" S: (n) Caterpillar, cat (a large tracked vehicle that is propelled by two endless metal belts; frequently used for moving earth in construction and farm work) S: (n) big cat, cat (any of several large cats typically able to roar and living in the wild) S: (n) computerized tomography, computed tomography, CT, computerized axial tomography, computed axial tomography, CAT (a method of examining body organs by scanning them with X rays and using a computer to construct a series of crosssectional scans along a single axis) S: (n) domestic cat, house cat, Felis domesticus, Felis catus (any domesticated member 13 of the genus Felis) 14 Wordnet 3.0 statistics POS Unique Synsets Total Word-Sense Pairs Strings Noun 117,798 82,115 146,312 Verb 11,529 13,767 25,047 Adjective 21,479 18,156 30,002 4,481 3,621 5,580 155,287 117,659 206,941 Adverb Totals 15 Wordnet 3.0 statistics POS Noun Verb Adjective Adverb Totals Monosemous Polysemous Polysemous Words and Senses Words Senses 101,863 15,935 44,449 6,277 5,252 18,770 16,503 4,976 14,399 3,748 733 1,832 128,391 26,896 79,450 16 Wordnet 3.0 statistics POS Average Polysemy Average Polysemy Including Monosemous Words Excluding Monosemous Words Noun 1.24 2.79 Verb 2.17 3.57 1.4 2.71 1.25 2.5 Adjective Adverb 17 http://www.visuwords.com 18 19 Usage of Wordnet • Improve recall of textual based analysis: – Query -> Index • • • • • Synonyms: commence – begin Hypernyms: taxi -> car Hyponyms: car -> taxi Meronyms: trunk -> elephant Lexical entailments: gun -> shoot • Inferencing: – what things can burn? • Expression in language generation and translation: – alternative words and paraphrases 20 Improve recall • Information retrieval: – small databases without redundancy, e.g. image captions, video text • Text classification: – small training sets • Question & Answer systems – query analysis: who, whom, where, what, when 21 Improve recall • Anaphora resolution: – The girl fell off the table. She.... – The glass fell of the table. It... • Coreference resolution: – When he moved the furniture, the antique table got damaged. • Information extraction (unstructed text to structured databases): – generic forms or patterns "vehicle" - > text with specific cases "car" 22 Improve recall • Summarizers: – Sentence selection based on word counts -> concept counts – Avoid repetition in summary -> language generation • Limited inferencing: detect locations, organisations, etc. 23 Many others • Data sparseness for machine learning: hapaxes can be replaced by semantic classes • Use redundancy for more robustness: spelling correction and speech recognition can built semantic expectations using Wordnet and make better choices • Sentiment and opinion mining • Natural language learning 24 Recall & Precision “jail” “nerve cell” “police cell” “neuron” found query: “cell” “cell phone” intersection “mobile phones” relevant recall = doorsnede / relevant Recall < 20% for basic search engines! precision = doorsnede / gevonden (Blair & Maron 1985) EuroWordNet • The development of a multilingual database with wordnets for several European languages • Funded by the European Commission, DG XIII, Luxembourg as projects LE2-4003 and LE4-8328 • March 1996 - September 1999 • 2.5 Million EURO. • http://www.hum.uva.nl/~ewn • http://www.illc.uva.nl/EuroWordNet/finalresultsewn.html 26 EuroWordNet • Languages covered: – EuroWordNet-1 (LE2-4003): English, Dutch, Spanish, Italian – EuroWordNet-2 (LE4-8328): German, French, Czech, Estonian. • Size of vocabulary: – EuroWordNet-1: 30,000 concepts - 50,000 word meanings. – EuroWordNet-2: 15,000 concepts- 25,000 word meaning. • Type of vocabulary: – the most frequent words of the languages – all concepts needed to relate more specific concepts 27 EuroWordNet Model Domains move go Air bewegen gaan 2OrderEntity Traffic III ride Ontology Location Dynamic Road` III rijden drive I III I II III II Lexical Items Table Lexical Items Table Lexical Items Table ILI-record {drive} Lexical Items Table III III II cabalgar jinetear II conducir III mover transitar berijden cavalcare guidare Inter-Lingual-Index I = Language Independent link II = Link from Language Specific to Inter lingual Index III = Language Dependent Link III andare muoversi 28 Differences in relations between EuroWordNet and WordNet • Added Features to relations • Cross-Part-Of-Speech relations • New relations to differentiate shallow hierarchies • New interpretations of relations 30 EWN Relationship Labels {airplane} HAS_MERO_PART: conj1 HAS_MERO_PART: conj2 disj1 HAS_MERO_PART: conj2 disj2 {door} {jet engine} {propeller} {door} HAS_HOLO_PART: disj1 HAS_HOLO_PART: disj2 HAS_HOLO_PART: disj3 {car} {room} {entrance} {dog} HAS_HYPERONYM: conj1 HAS_HYPERONYM: conj2 {mammal} {pet} {albino} HAS_HYPERONYM: disj1 HAS_HYPERONYM: disj2 {plant} {animal} Default Interpretation: non-exclusive disjunction 32 EWN Relationship Labels Factive/Non-factive CAUSES (Lyons 1977) factive (default interpretation): “to kill causes to die”: {kill} CAUSES {die} non-factive: E1 probably or likely causes event E2 or E1 is intended to cause some event E2: “to search may cause to find”. {search} CAUSES {find} non-factive 33 Cross-Part-Of-Speech relations WordNet1.5: nouns and verbs are not interrelated by basic semantic relations such as hyponymy and synonymy: adornment 2 adorn 1 change of state-- (the act of changing something) change, alter-- (cause to change; make different) EuroWordNet: words of different parts of speech can be inter-linked with explicit xpos-synonymy, xpos-antonymy and xpos-hyponymy relations: {adorn V} {size N} XPOS_NEAR_SYNONYM XPOS_NEAR_HYPONYM {adornment N} {tall A} {short A} 34 Role relations In the case of many verbs and nouns the most salient relation is not the hyperonym but the relation between the event and the involved participants. These relations are expressed as follows: {knife} {to cut} {school} {to teach} ROLE_INSTRUMENT INVOLVED_INSTRUMENT ROLE_LOCATION INVOLVED_LOCATION {to cut} {knife} {to teach} {school} reversed reversed These relations are typically used when other relations, mainly hyponymy, do not clarify the position of the concept network, but the word is still closely related to another word. 35 Co_Role relations guitar player player to play music guitar ice saw saw ice HAS_HYPERONYM CO_AGENT_INSTRUMENT HAS_HYPERONYM ROLE_AGENT CO_AGENT_INSTRUMENT HAS_HYPERONYM ROLE_INSTRUMENT HAS_HYPERONYM CO_INSTRUMENT_AGENT HAS_HYPERONYM CO_INSTRUMENT_PATIENT HAS_HYPERONYM ROLE_INSTRUMENT CO_PATIENT_INSTRUMENT player guitar person to play music musical instrument to make musical instrument musical instrument guitar player saw ice saw to saw ice saw REVERSED 36 Co_Role relations Examples of the other relations are: criminal novel writer/ poet dough photograpic camera CO_AGENT_PATIENT CO_AGENT_RESULT CO_PATIENT_RESULT CO_INSTRUMENT_RESULT victim novel/ poem pastry/ bread photo 37 Overview of the Language Internal relations in EuroWordnet Same Part of Speech relations: NEAR_SYNONYMY HYPERONYMY/HYPONYMY ANTONYMY HOLONYMY/MERONYMY apparatus - machine car - vehicle open - close head - nose Cross-Part-of-Speech relations: XPOS_NEAR_SYNONYMY dead - death; to adorn - adornment XPOS_HYPERONYMY/HYPONYMY to love - emotion XPOS_ANTONYMY to live - dead CAUSE die - death SUBEVENT buy - pay; sleep - snore ROLE/INVOLVED write - pencil; hammer - hammer STATE the poor - poor MANNER to slurp - noisily 38 BELONG_TO_CLASS Rome - city Horizontal & vertical semantic relations chronical patient ; mental patient HYPONYM ρ-PATIENT patient STATE cure ρ-CAUSE docter treat ρ-PATIENT ρ-AGENT HYPONYM child docter disease; disorder HYPONYM stomach disease, kidney disorder, ρ-PROCEDURE physiotherapy medicine etc. ρ-LOCATION co-ρAGENT-PATIENT hospital, etc. child The Multilingual Design • Inter-Lingual-Index: unstructured fund of concepts to provide an efficient mapping across the languages; • Index-records are mainly based on WordNet synsets and consist of synonyms, glosses and source references; • Various types of complex equivalence relations are distinguished; • Equivalence relations from synsets to index records: not on a word-to-word basis; • Indirect matching of synsets linked to the same index items; 40 Equivalent Near Synonym 1. Multiple Targets (1:many) Dutch wordnet: schoonmaken (to clean) matches with 4 senses of clean in WordNet1.5: • make clean by removing dirt, filth, or unwanted substances from • remove unwanted substances from, such as feathers or pits, as of chickens or fruit • remove in making clean; "Clean the spots off the rug" • remove unwanted substances from - (as in chemistry) 2. Multiple Sources (many:1) Dutch wordnet: versiersel near_synonym versiering ILI-Record: decoration. 3. Multiple Targets and Sources (many:many) Dutch wordnet: toestel near_synonym apparaat ILI-records: machine; device; apparatus; tool 41 Equivalent Hyperonymy Typically used for gaps in English WordNet: • genuine, cultural gaps for things not known in English culture: – Dutch: klunen, to walk on skates over land from one frozen water to the other • pragmatic, in the sense that the concept is known but is not expressed by a single lexicalized form in English: – Dutch: kunststof = artifact substance <=> artifact object 42 Equivalent Hyponymy has_eq_hyponym Used when wordnet1.5 only provides more narrow terms. In this case there can only be a pragmatic difference, not a genuine cultural gap, e.g.: Spanish dedo = either finger or toe. 43 Complex mappings across languages EN-Net IT-Net toe dito finger { toe : part of foot } head { finger : part of hand } { dedo , dito : finger or toe } { head : part of body } NL-Net hoofd kop { hoofd : human head } { kop : animal head } ES-Net dedo = normal equivalence = eq _has_hyponym = eq _has_hyperonym 44 Typical gaps in the (English) ILI • Dutch: doodschoppen (to kick to death): eq_hyperonym {kill}V and to {kick}V aardig (Adjective, to like): eq_near_synonym {like}V cassière (female cashier) eq_hyperonym {cashier}, {woman} kunstproduct (artifact substance) eq_hyperonym {artifact} and to {product} • Spanish: alevín (young fish): eq_hyperonym {fish} and eq_be_in_state {young} cajera (female cashier) eq_hyperonym {cashier}, {woman} 45 Wordnets as semantic structures • Wordnets are unique language-specific structures: – – – – different lexicalizations differences in synonymy and homonymy different relations between synsets same organizational principles: synset structure and same set of semantic relations. • Language independent knowledge is assigned to the ILI and can thus be shared for all language linked to the ILI: both an ontology and domain hierarchy 46 Autonomous & Language-Specific Wordnet1.5 Dutch Wordnet voorwerp {object} object artifact, artefact (a man-made object) block natural object (an object occurring naturally) blok {block} instrumentality body implement lichaam {body} device container tool instrument box werktuig{tool} spoon bag bak {box} lepel {spoon} tas {bag} 47 Linguistic versus Artificial Ontologies Artificial ontology: • better control or performance, or a more compact and coherent structure. • introduce artificial levels for concepts which are not lexicalized in a language (e.g. instrumentality, hand tool), • neglect levels which are lexicalized but not relevant for the purpose of the ontology (e.g. tableware, silverware, merchandise). What properties can we infer for spoons? spoon -> container; artifact; hand tool; object; made of metal or plastic; for eating, pouring or cooking 48 Linguistic versus Artificial Ontologies Linguistic ontology: • Exactly reflects the relations between all the lexicalized words and expressions in a language. • Captures valuable information about the lexical capacity of languages: what is the available fund of words and expressions in a language. What words can be used to name spoons? spoon -> object, tableware, silverware, merchandise, cutlery, 49 Wordnets versus ontologies • Wordnets: • autonomous language-specific lexicalization patterns in a relational network. • Usage: to predict substitution in text for information retrieval, • text generation, machine translation, wordsense-disambiguation. • Ontologies: • data structure with formally defined concepts. • Usage: making semantic inferences. 50 Sharing world knowledge • All wordnets in the world can be linked to the same ontology • All wordnets in the world can be linked to the same thesaurus 51 Wordnet: Domain information Concepts Vocabularies of languages 1 2 bank rec: 12345 - financial institute Clothing rec: 54321 - river side rec: 9876 - small string instrument 2 rec: 65438 - musician playing a violin violist rec:42654 - musician 1 2 Domains Culture Sport Finance Music 1 violin string Relations Ball Winter sports sports type-of rec:35576 - string of an instrument part-of type-of rec:29551 - underwear rec:25876 - string instrument 52 How to harmonize wordnets? • Wordnets are unique language-specific lexicalizations patterns • Define universal sets of concepts that play a major role in many different wordnets: so-called Base Concepts • Define base concepts in each language wordnet – High level in the hierarchy – Many hyponyms • Provide the closest equivalent in English wordnet • Determine the intersection of English equivalences 53 Lexicalization patterns entity object garbage threat artifact organism animal building bird church canary dog crocodile plant tree flower rose 25 unique beginners 1024 base concepts basic level concepts abbey common canary 54 Base Concept Intersection Nouns Verbs Intersection EN, NL, IT, ES 24 6 Intersection FR, DE, EE, CZ 70 30 Intersection All 13 2 {human 1; individual#1; mortal#1; person#1; someone#1; soul#1} {animal 1; animate being#1; beast#1; brute#1; creature#1; fauna#1} {flora 1; plant#1; plant life#1} {matter 1; substance#1} {food 1; nutrient#1} {feeling 1} {act 1; human action#1; human activity#1} {cause 6; get#9; have#7; induce#2; make#12; stimulate#3} {create 2; make#13} {go 14; locomote#1; move#15; travel#4} {be 4; have the quality of being#1} 55 Explanations for low intersection of Base Concepts • The individual selections are not representative enough. • There are major differences in the way meanings are classified, which have an effect on the frequency of the relations. • The translations of the selection to WordNet1.5 synsets are not reliable • The resources cover very different vocabularies 56 Concepts selected by at least two languages: intersections of pairs NOUNS NL ES IT EN VERBS NL ES IT EN NL ES IT EN 1027 103 182 333 323 36 42 86 103 523 45 284 36 128 18 43 182 45 334 167 42 18 104 39 333 284 167 1296 86 43 39 236 57 Common Base Concepts Nouns Verbs Physical objects & substances 491 Processes and states 272 Mental objects Total Total 491 228 33 796 500 33 228 1024 58 Table 4: Number of Common BCs represented in the local wordnets NL ES IT Related to CBCs Eq_synonym Eq_near 992 1012 878 269 0 191 725 1009 759 CBCs Without Direct Equivalent 97 15 9 Table 5: BC4 Gaps in at least two wordnets (10 synsets) body covering#1 mental object#1; cognitive content#1; content#2 body substance#1 natural object#1 social control#1 place of business#1; business establishment#1 change of magnitude#1 plant organ#1 contractile organ#1 plant part#1 psychological feature#1 spatial property#1; spatiality#1 59 Table 6: Local senses with complex equivalence relations to CBCs Eq_has_hyperonym eq_has_hyponym Eq_has_holonym Eq_has_meronym Eq_involved Eq_is_caused_by Eq_is_state_of NL 61 34 2 3 3 3 1 ES 40 14 0 2 IT 4 20 Example of complex relation CBC: cause to feel unwell#1, Verb Closest Dutch concept: {onwel#1}, Adjective (sick) Equivalence relation: eq_is_caused_by 60 EuroWordNet data Dutch Spanish Italian French German Czech Estonian English WN15 Synsets No. of senses Sens./ Entries Sens./ LIRels. LIRels/ EQRels- EQRels/s Synsets syns. entry syns ILI yn without ILI 44015 70201 1,59 56283 1,25 111639 2,54 53448 1,21 7203 23370 50526 2,16 27933 1,81 55163 2,36 21236 0,91 0 40428 48499 1,20 32978 1,47 117068 2,90 71789 1,78 1561 22745 32809 1.44 18777 1.75 49494 2.18 22730 1.00 20 15132 20453 1.35 17098 1.20 34818 2.30 16347 1.08 0 12824 19949 1.56 12283 1.62 26259 2.05 12824 1.00 0 7678 13839 1.80 10961 1.26 16318 2.13 9004 1.17 0 16361 40588 2,48 17320 2,34 42140 2,58 n.a. n.a. n.a. 94515 187602 1,98 126617 1,48 211375 2,24 n.a. n.a. n.a. 61 From EuroWordNet to Global WordNet • Currently, wordnets exist for more than 50 languages, including: • Arabic, Bantu, Basque, Chinese, Bulgarian, Estonian, Hebrew, Icelandic, Japanese, Kannada, Korean, Latvian, Nepali, Persian, Romanian, Sanskrit, Tamil, Thai, Turkish, Zulu... • Many languages are genetically and typologically unrelated • http://www.globalwordnet.org 62 Global Wordnet Association EuroWordNet • • • • • • • • English German Spanish French Italian Dutch Czech Estonian BalkaNet Romanian Bulgarian Turkish Slovenian Greek Serbian http://www.globalwordnet.org • • • • • • • • • Danish Norway Swedish Portuguese Korean Russian Basque Catalan Thai Arabic Polish Welsh Chinese 20 Indian Languages Brazilian Portuguese Hebrew Latvian Persian Kurdish Avestan Baluchi Hungarian 63 Some downsides of the EuroWordnet model • Construction is not done uniformly • Coverage differs • Not all wordnets can communicate with one another • Proprietary rights restrict free access and usage • A lot of semantics is duplicated • Complex and obscure equivalence relations due to linguistic differences between English and other languages 64 Next step: Global WordNet Grid Fahrzeug 1 Auto Zug Inter-Lingual Ontology vehicle voertuig 1 auto trein 1 car Object train 2 Dutch Words 2 English Words TransportDevice vehículo 1 véhicule veicolo voiture 1 auto treno 2 Italian Words dopravní prostředník 1 auto vlak 2 Czech Words liiklusvahend auto killavoor 3 auto tren Spanish Words 1 Device 3 2 2 German Words 2 Estonian Words 1 train 2 French Words 65 GWNG: Main Features • Construct separate wordnets for each Grid language • Contributors from each language encode the same core set of concepts plus culture/language-specific ones • Synsets (concepts) can be mapped crosslinguistically via an ontology 66 The Ontology: Main Features • Formal ontology serves as universal index of concepts • List of concepts is not just based on the lexicon of a particular language (unlike in EuroWordNet) but uses ontological observations • Ontology contains only upper and mid-level concepts • Concepts are related in a type hierarchy • Concepts are defined with axioms 67 The Ontology: Main Features • In addition to high-level (“primitive”) concept ontology needs to express low-level concepts lexicalized in the Grid languages • Additional concepts can be defined with expressions in Knowledge Interchange Format (KIF) based on first order predicate calculus and atomic element 68 The Ontology: Main Features • Minimal set of concepts (Reductionist view): – to express equivalence across languages – to support inferencing • Ontology must be powerful enough to encode all concepts that are lexically expressed in any of the Grid languages • Ontology need not and cannot provide a linguistic encoding for all concepts found in the Grid languages – Lexicalization in a language is not sufficient to warrant inclusion in the ontology – Lexicalization in all or many languages may be sufficient • Ontological observations will be used to define the concepts in the ontology 69 Ontological observations • Identity criteria as used in OntoClean (Guarino & Welty 2002), : – rigidity: to what extent are properties true for entities in all worlds? You are always a human, but you can be a student for a short while. – essence: what properties are essential for an entity? Shape is essential for a statue but not for the clay it is made of. – unicity: what represents a whole and what entities are parts of these wholes? An ocean is a whole but the water it contains is not. 70 Type-role distinction • Current WordNet treatment: (1) a husky is a kind of dog(type) (2) a husky is a kind of working dog (role) • What’s wrong? (2) is defeasible, (1) is not: *This husky is not a dog This husky is not a working dog Other roles: watchdog, sheepdog, herding dog, lapdog, etc…. 71 Ontology and lexicon •Hierarchy of disjunct types: Canine PoodleDog; NewfoundlandDog; GermanShepherdDog; Husky •Lexicon: – NAMES for TYPES: {poodle}EN, {poedel}NL, {pudoru}JP ((instance x Poodle) – LABELS for ROLES: {watchdog}EN, {waakhond}NL, {banken}JP ((instance x Canine) and (role x GuardingProcess)) 72 Ontology and lexicon •Hierarchy of disjunct types: River; Clay; etc… •Lexicon: – NAMES for TYPES: {river}EN, {rivier, stroom}NL ((instance x River) – LABELS for dependent concepts: {rivierwater}NL (water from a river => water is not a unit) {kleibrok}NL (irregularly shared piece of clay=>non-essential) ((instance x water) and (instance y River) and (portion x y) ((instance x Object) and (instance y Clay) and (portion x y) and (shape X Irregular)) 73 Rigidity • The “primitive” concepts represented in the ontology are rigid types • Entities with non-rigid properties will be represented with KIF statements • But: ontology may include some universal, core concepts referring to roles like father, mother 74 Properties of the Ontology • Minimal: terms are distinguished by essential properties only • Comprehensive: includes all distinct concepts types of all Grid languages • Allows definitions via KIF of all lexemes that express non-rigid, non-essential properties of types • Logically valid, allows inferencing 75 Mapping Grid Languages onto the Ontology • Explicit and precise equivalence relations among synsets in different languages: – type hierarchy is minimal – subtle differences can be encoded in KIF expressions • Grid database contains wordnets with synsets that label • --either “primitive” types in the hierarchies, • --or words relating to these types in ways made explicit in KIF expressions • If 2 lgs. create the same KIF expression, this is a statement of equivalence! 76 How to construct the GWNG • Take an existing ontology as starting point; • Use English WordNet to maximize the number of disjunct types in the ontology; • Link English WordNet synsets as names to the disjunct types; • Provide KIF expressions for all other English words and synsets • Copy the relation to the ontology to other languages, including KIF statements built for English • Revise KIF statements to make the mapping more precise • Map all words and synsets that are and cannot be mapped to English WordNet to the ontology: – propose extensions to the type hierarchy – create KIF expressions for all non-rigid concepts 77 Initial Ontology: SUMO (Niles and Pease) SUMO = Suggested Upper Merged Ontology --consistent with good ontological practice --fully mapped to WordNet(s): 1000 equivalence mappings, the rest through subsumption --freely and publicly available --allows data interoperability --allows NLP --allows reasoning/inferencing 78 SUMO • 1,000 generic, abstract, high-level terms • 4,000 definitional statements • MILO (Mid-Level Ontology) closer to lexicon, WordNet 79 Mapping Grid languages onto the Ontology • Check existing SUMO mappings to Princeton WordNet -> extend the ontology with rigid types for specific concepts • Extend it to many other WordNet synsets • Observe OntoClean principles! (Synsets referring to non-rigid, non-essential, nonunicitous concepts must be expressed in KIF) 80 Lexicalizations not mapped to WordNet • Not added to the type hierarchy: {straathond}NL (a dog that lives in the streets) ((instance x Canine) and (habitat x Street)) • Added to the type hierarchy: {klunen}NL (to walk on skates from one frozen body to the next over land) WalkProcess KluunProcess Axioms: (and (instance x Human) (instance y Walk) (instance z Skates) (wear x z) (instance s1 Skate) (instance s2 Skate) (before s1 y) (before y s2) etc… • National dishes, customs, games,.... 81 Most mismatching concepts are not new types • Refer to sets of types in specific circumstances or to concept that are dependent on these types, next to {rivierwater}NL there are many other: {theewater}NL (water used for making tea) {koffiewater}NL (water used for making coffee) {bluswater}NL (water used for making extinguishing file) • Relate to linguistic phenomena: – gender, perspective, aspect, diminutives, politeness, pejoratives, part-of-speech constraints 82 KIF expression for gender marking • {teacher}EN ((instance x Human) and (agent x TeachingProcess)) • {Lehrer}DE ((instance x Man) and (agent x TeachingProcess)) • {Lehrerin}DE ((instance x Woman) and (agent x TeachingProcess)) 83 KIF expression for perspective sell: subj(x), direct obj(z),indirect obj(y) versus buy: subj(y), direct obj(z),indirect obj(x) (and (instance x Human)(instance y Human) (instance z Entity) (instance e FinancialTransaction) (source x e) (destination y e) (patient e) The same process but a different perspective by subject and object realization: marry in Russian two verbs, apprendre in French can mean teach and learn 84 Aspectual variants • Slavic languages: two members of a verb pair for an ongoing event and a completed event. • English: can mark perfectivity with particles, as in the phrasal verbs eat up and read through. • Romance languages: mark aspect by verb conjugations on the same verb. • Dutch, verbs with marked aspect can be created by prefixing a verb with door: doorademen, dooreten, doorfietsen, doorlezen, doorpraten (continue to breathe/eat/bike/read/talk). • These verbs are restrictions on phases of the same process • Does NOT warrant the extension of the ontology with separate processes for each aspectual variant 85 Kinship relations in Arabic • • • • عم father's brother, َ (Eam~) paternal uncle. ( خَالxaAl) mother's brother, maternal uncle. ع َّمة َ (Eam~ap) father's sister, paternal aunt. ( خَالَةxaAlap) mother's sister, maternal aunt 86 Kinship relations in Arabic • • • • ......... ش ِقيقَة َ ($aqiyqapfull) sister, sister on the paternal and maternal side (as distinct from >( أ ُ ْختuxot): 'sister' which may refer to a 'sister' from paternal or maternal side, or both sides). ( ثَ ْكالنvakolAna) father bereaved of a child (as opposed to ( يَ ِتيمyatiym) or ( يَ ِتي َمةyatiymap) for feminine: 'orphan' a person whose father or mother died or both father and mother died). ( ثَ ْكلَىvakolaYa) other bereaved of a child (as opposed to يَتِيمor يَتِي َمةfor feminine: 'orphan' a person whose father or mother died or both father and mother died). 87 Complex Kinship concepts father's brother, paternal uncle WORDNET paternal uncle => uncle => brother of ....???? ONTOLOGY (=> (paternalUncle ?P ?UNC) (exists (?F) (and (father ?P ?F) (brother ?F ?UNC)))) 88 Universality as evidence • English verb cut abstracts from the precise process but there are troponyms that implicate the manner : – snip, clip imply scissors, chop and hack a large knife or an axe • Dutch there is no general verb but only specific verbs: knippen “clip, snip, cut with scissors or a scissor-like tool'”, snijden “cut with a knife or knife-like tool”, hakken “chop, hack, to cut with an axe, or similar tool”). • If lexicalization of the specific process is more universal it can be seen as evidence that the specific processes should be listed in the ontology and not the generic verb 89 Open Questions/Challenges • What is a word, i.e., a lexical unit? • What is the status of complex lexemes like English lightning rod, word of mouth, find out, kick the bucket? • What is a semantic unit, i.e. a concept? 90 Open Questions/Challenges • Is there a core inventory of concepts that are universally encoded? • If so, what are these concepts? • How can crosslinguistic equivalence be verified? • Is there systematicity to the language-specific extensions? • What are the lexicalization patterns of individual languages? • Are lexical gaps accidental or systematic? 91 Coverage: what belongs in a universal lexical database? • Formal, linguistic criteria for inclusion • Informal, cultural criteria • Both are difficult to define and apply! 92 Advantages of the Global Wordnet Grid • Shared and uniform world knowledge: – universal inferencing – uniform text analysis and interpretation • More compact and less redundant databases • More clear notion how languages map to the knowledge – better criteria for expressing knowledge – better criteria for understanding variation 93 Expansion with pure hyponymy relations dog hunting dog puppy dachshund lapdog street dog poodle bitch watchdog short hair dachshund long hair dachshund Expansion from a type to roles 94 Expansion with pure hyponymy relations dog hunting dog puppy dachshund lapdog street dog poodle bitch watchdog short hair dachshund long hair dachshund Expansion from a role to types and other roles 95 Automotive ontology: (http://www.ontoprise.de) 96 Who uses ontologies? 97 98