Extending the SpanishWordNet Clara Soler Universitat Ramon Llull
Jan 31, 2016
Extending the SpanishWordNet
Clara SolerUniversitat Ramon Llull
Extending the SpanishWordNet
● Aims:– Proposal about:
A new classification and a new way of representing Adjectives and their polysemy in the lexical database
● By means of:– Using MikroKosmos Ontology (Nirenburg 1994
et al.) to modificate if necessary EuroWordNet Top Concept Ontology
Index● Classification of Adjectives in WordNet● How Polysemy of Adjectives is represented● Classification of Adjectives in EuroWordNet –
SpanishWordNet● The Ontological Criteria ● Mikrokosmos Ontology / Treatment of Adjective
Polysemy● EuroWordNet Top Ontology● Conclusions
Classification of Adjectives in WordNet
● WordNet 1.5 contains 16428 adjectivals synsets (including nouns, participles and prepositional phrases which work as adjectives)
● Kept in three files: Descriptive, Relationals and Participles.
● How are they organized? Each class expresses typical syntactic and semantic features.
● Antonymy is the lexical relation that divides adjectives: Descriptives adjectives have antonyms and Relationals do not.
Classification of Adjectives in WN
● Descriptives adjectives are organized into non-hierarchic synsets formed by one, or more, pairs of antonyms. These synsets are clusters that represent values, from one extreme to the other, of some attribute.
● Relational adjectives are represented with pointers to the noun or verb from which they derive (a pointer expresses a lexical relation)
● Participle adjectives are considered a kind of Descriptives adjectives ending in -ing and -ed, they do not suit very much in cluster structures and are kept in a separated file.
Classification of Adjectives in WN● Descriptives adjectives: attribute Warmth
torrid – hot – warm – tepid – cool – cold- frigid● Relational adjectives
2 senses of cosmologic Sense 1cosmologic, cosmological, cosmogonic, cosmogonical, cosmogenic, cosmogenical -- (pertaining to the branch of astronomy dealing with the origin and history and structure and dynamics of the universe; "cosmologic science"; "cosmological redshift"; "cosmogonic theories of the origin of the universe")Pertains to noun cosmology (Sense 2)=>cosmology, cosmogony, cosmogeny -- (the branch of astrophysics that studies the origin and evolution and structure of the universe)
Classification of Adjectives in WN● Relational adjectives
Sense 2cosmologic, cosmological -- (pertaining to the branch of philosophy dealing with the elements and laws and especially the characteristics of the universe such as space and time and causality; "cosmologic philosophy"; "a cosmological argument is an argument that the universe demands the admission of an adequate external cause which is God")Pertains to noun cosmology (Sense 1)=>cosmology -- (the metaphysical study of the origin and nature of the universe)=> metaphysics -- (the philosophical study of being and knowing)
How Polysemy of Adjectives is represented
● WordNet understands polysemy as a correlation of the frequency of occurrence ( Zipf 1945). It means that on the average, the more frequently a word is used the more different meanings it will have.
● Therefore, new, which is :– Highly frequent, it is considered– Highly polysemous
How Polysemy of Adjectives is represented
● 1. new (vs. old) -- (not of long duration; having just (or relatively recently) come into being or been made or acquired or discovered; "a new law"; "new cars"; "a new comet"; "a new friend"; "a new year"; "the New World")2. new(prenominal) -- (other than the former one(s); different; "they now have a new leaders"; "my new car is four years old but has only 15,000 miles on it"; "ready to take a new direction")3. new, unexampled -- (having no previous example or precedent or parallel; "a time of unexampled prosperity")
How Polysemy of Adjectives is represented
● 4. fresh, new, novel -- (of a kind not seen before; "the computer produced a completely novel proof of a well-known theorem")5. raw, new, wet behind the ears(predicate) -- (lacking training or experience; "the new men were eager to fight"; "raw recruits"; "he was still wet behind the ears when he shipped as a hand on a merchant vessel")6. newfangled, new -- (of a new (often outrageous) kind or fashion)
7. new, new to(predicate) -- ((often followed by `to') unfamiliar; "new experiences"; "experiences new to him"; "errors of someone new to the job")
How Polysemy of Adjectives is represented
● 8. new, young -- ((of crops) harvested at an early stage of development; before complete maturity; "new potatoes"; "young corn")9. new -- (unaffected by use or exposure; "it looks like new")10. New -- (in use after Medieval times; "New Eqyptian was the language of the 18th to 21st dynasties")11. Modern, New -- (used of a living language; being the current stage in its development; "Modern English"; "New Hebrew is Israeli Hebrew")
Classification of Adjectives in EuroWordNet
● Adjectives were not included in EuroWordNet, because:– Being a modifier, the information that conveys is
less vital– It was considered a highly polysemous category-
therefore representation in an enumerative lexicon becomes difficult
Classification of Adjectives in the SpanishWordNet
● Spanish and Catalan adjectives are translations of the English adjectives included in WordNet 1.5.
● They are classified and represented in the same way
Classification of Adjectives in the SWN● 01256444a
all lock 0 new_1 0 nuevo_5 0 nou_1 not of long duration; having just come into being or been made or acquired or discovered: "a new law"; "new cars"; "a new comet"; "a new friend"
● 01290120a all lock 0 fresh_2 novel_1 new_2 0 fresco_8 nuevo_8 novel_1 0 fresc_2 novell_1 nou_2 of a kind not seen before: "the computer produced a completely novel proof of a well-known theorem"
Classification of Adjectives in the SWN
● 00104365a all lock 0 new_4 unexampled_1 lock 0 inaudito_1 nuevo_2 lock 0 inaudit_2 nou_4 "an unprecedented expansion in population and industry"
00803732a all lock 0 hot_12 new_5 0 fresco_4 nolex 0 "hot off the press"; "fresh ideas"
Classification of Adjectives in the SWN● all lock 0 new_8 newfangled_1
lock 0 novedoso_1 nolex 0 of a new (often outrageous) kind or fashion
● all lock 0 modern_5 new_9 lock 0 moderno_2 contemporáneo_6 lock 0 modern_4 contemporani_4 used of a living language; being the current stage in its development: "Modern English"; "New Hebrew is Israeli Hebrew" Perteneciente o relativo al tiempo actual Pertanyent o relatiu al temps actual
Extending the SpanishWordNet
● At issue: – Relationals are not polysemous?– It is possible to represent polysemy of Adjectives
without enumerating all the possible shades they may take?
– Is it possible to relate polysemy with other parameters differents from frequency, compatibility with greater or smaller number of nominals, noun desambiguation?
– Are Adjectives that polysemous?
Extending the SpanishWordNet
● This proposal bases the extending of the SpanishWordnet by incorporating Adjectives on the following premises:– Adjectives will be classified too in Descriptive
and Relationals notwithstanding based on ontological criteria.
– Their polysemy will also be accounted by the same ontological taxonomy.
– They can also be classified according to the Base Concepts of the Top Ontology.
The Ontological Criteria
● These criteria come from within the framework of the Mikrokosmos Ontology (Nirenburg 1994 et al.).
● In this model the lexicon mediates between a language of Text Meaning Representation and an ontology.
● Adjective meaning is explained according to this conceptual ontology.
● Lexical entries are instances of ontological types.
The MikroKosmos Ontology
● Classification of Adjectives:– Scalars:
● based on Property ontological concepts Numerical scale : Warmth
torrid – hot – warm – tepid – cool – cold - frigid
Literal scale : colour
red – yellow – black – white – blue – orange- ● Evaluative
Terrific – Great – Good – Regular – Bad – Awful - Horrible
The MikroKosmos Ontology● Non-Scalars (Relationals):
● Denominals: Object ontological concepts National – nation● Deverbals: Event ontological concepts Readable – read
● An Adjectival Class: an ontological type
The MikroKosmos Ontology● Numerical scale properties:
– Evaluation Price– Size Speed– Mass Ease– Girth– Age– Cost
● Literal scale properties:– Orientation Direction– Side Color– Shape
Treatment of Adjectival Polysemy● In this proposal:
– Polysemy of a lexical item implies a change of meaning
● A romantic style / a romantic adventure● In Spanish: negocio familiar / una atmosfera familiar● In English: a family business / a familiar atmosphere● Good son / Good deal : always positive
– Polysemy has to do with relatedness of senses arising through extension of meaning (Aarts & Calbert, 1979)
– This relatedness of senses can be explained by the ontological model
Treatment of Adjectival Polysemy● A change of Ontological Type: from an Ontology of
Entities or Events to an Ontology of Properties– Victor Hugo is a French romantic author who
lived during the XIX century– Venice is a very romantic city for lovers
● A change within the same Ontological Type: – High altitude: Scale Height – High ideals: Scale Evaluative
EuroWordNet Top Concept Ontology
● Top Concept Ontology classifies the Base Concepts
● Base Concepts are the most important prevailing meanings in the local wordnets, and constitute a shared set of 1059. Represent the shared cores of the different wordnets.
● Top Concept Ontology consists of 63 fundamental semantic distinctions, organized by subtype and opposition relations, which organize the Base Concepts
EuroWordNet Top Concept Ontology
● It is linked to the Inter-Lingual-Index (ILI) and provides some language-independent structuring of it.
● The 63 semantic distinctions or Top Concepts are classified according to three types of entities: 1st-Order-Entities, 2nd-Order-Entities and 3rd-Order-Entities.
EuroWordNet Top Concept Ontology
● 1st-Order-Entities: expressed by concrete nouns, concrete entities. One euro, a college, a hobbit
● 3rd-Order-Entities: expressed by abstract nouns: any unobservable proposition. Platonic ideas
EuroWordNet Top Concept Ontology ● 2nd-Order-Entities:
– denoted by any part of speech (nouns, verbs, adjectives and adverbs).
– represent any static or dynamic situation that cannot be grasped, seen, felt, or experienced as an independent physical thing.
– located in time, they happen, rather than exist.
EuroWordNet Top Concept Ontology● 2ndOrderEntities: (two classification schemes)
– Situation Type: the event-structure in terms of which a situation can be characterized as a conceptual unit over time
● Dynamic: transition-BoundedEvent-UnboundedEvent
● Static-Property: colour, speed, age, length, size, shape, weight. (single entity)
-Relation: relation, kinship, distance, space (two entities)
EuroWordNet Top Concept Ontology
– Usage– Time– Social– Quantity– Purpose– Possession– Physical
Situation Components (BC as conceptually coherent clusters. SC reflect the most salient semantic components)
EuroWordNet Top Concept Ontology
● Modal● Mental● Manner● Location● Experience● Existence● Condition● Communication
EuroWordNet Top Concept Ontology
● Cause● Agentive● Stimulating● Phenomenal
Proposal: to reuse Top Concepts as a new Property, Object and Event ontological concepts, and viceversa.
Extending the SpanishWordNet
● The idea is to use the ontological structure of MikroKosmos in events, objects and properties to introduce and classify the adjectives in the SpanishWordNet
● In the Top Ontology: to verify if all the Top Concepts Concepts of 2nd-Order-Entities denote all possible adjectives-situations
Extending the SpanishWordNet
● Joining both ontologies– Incorporation of the adjectival synsets into the
Inter-Lingual-Index– Incorporation of the new Top Concepts that
represent the adjectival scales into the class of 2nd-Order-Entities.
– Scalar adjectives will be defined according to the Base Concepts of 2nd-Order-Entities
– Non Scalar adjectives (Denominals and Deverbals) will be defined in relation with the Base Concepts of all Entities.
Extending the SpanishWordNet● Classification(s) of adjectives in the SWN:
– Following the MikroKosmos ontological taxonomy, adjectives can be classified according to the three types of entities:
● Scalars: subsumed under 2nd-Order-Entities● Non Scalars: subsumed under 1st, 2nd and 3rd
Order Entities (Denominals) and under 2nd-Order-Entities (Deverbals)
– 2nd -Order-Entities are denoted by adjectives. Therefore all of them can be classified according to the Base Concepts of these Entities
Extending the SpanishWordNet● Proposal of an adjectival synset which
represents polysemy: denominal and scalar● The adjective "romantic" has 3 senses in WordNet.
1. romantic, romanticist, romanticistic -- (belonging to or characteristic of romanticism or the Romantic movement in the arts; "romantic poetry")2. amatory, amorous, romantic -- (expressive of or exciting sexual love or romance; "her amatory affairs"; "amorous glances"; "a romantic adventure"; "a romantic moonlight ride")3. quixotic, romantic, wild-eyed -- (not sensible about practical matters; unrealistic; "as quixotic as a restoration of medieval knighthood"; "a romantic disregard for money"; "a wild-eyed dream of a world state")
Extending the SpanishWordNet● Proposal of an adjectival synset which represents
polysemy: denominal and scalar
The adjective romantic has also 3 senses in the SWN.1. romantic, romanticist, romanticistic
– Based on an Object ontological concept (Romanticism): denominal
2. romantic, amorous
– Based on a Property ontological concept (scale: evaluative (sentimentalism)
3. romantic, wild-eyed
– Based on a Property ontological concept (scale: evaluative (idealism)
Extending the SpanishWordNet● Proposal of an adjectival synset which represents
polysemy: scalar
The adjective new has only 4 senses in the SPW
1. new (vs. old) -- (not of long duration; having just (or relatively recently)
Based on a Property ontological concept (scale: temporal (duration)
2. new(prenominal) -- (other than the former one(s); different;)3. new, unexampled -- (having no previous example or precedent or parallel;)
4. fresh, new, novel -- (of a kind not seen before; )
Extending the SpanishWordNet● Proposal of an adjectival synset which represents
polysemy: scalar
5. raw, new, wet behind the ears(predicate) -- (lacking training or experience; )
6. newfangled, new -- (of a new (often outrageous) kind or fashion)
7. new, new to(predicate) -- ((often followed by `to') unfamiliar; )
9. new -- (unaffected by use or exposure; "it looks like new")
Based on a Property ontological concept (scale: newness)
Extending the SpanishWordNet● Proposal of an adjectival synset which represents
polysemy: scalar
8. new, young -- ((of crops) harvested at an early stage of development; before complete maturity;)Based on a Property ontological concept (scale: maturity)
11. Modern, New -- (used of a living language; being the current stage in its development; )Based on a Property ontological concept (scale: temporary nature)
Conclusions
● This paper tries to propose a new classification of Adjectives heading towards an extension of the SpanishWordNet, as well as a new way of representing polysemy.
● The following has been shown:– Relationals are also polysemous– Adjective polysemy can be reduced and
represented– Ontological criteria can reflect the relationship
between diferents senses of a lexical item
Conclusions
● Future work includes:– To focus on the syntagmatic level of the
Adjective-Noun combinations, and to study the diferent grades of co-ocurrence (Of compositive, semicompositive and noncompositive character).
– To treat if possible the called adverbials adjectives using the same ontological structure