LIN 6932 Lecture 10 Computational Lexical Semantics Hana

LIN 6932 Lecture 10: Computational Lexical Semantics Hana Filip 12/30/2021 LIN 6932 1

Meaning • Traditionally, meaning in language has been studied from three perspectives – The meanings of individual words – How those meanings combine to make meanings for individual sentences or utterances (compositional semantics) – How those meanings combine to make meanings for a text or discourse (discourse semantics) • We are going to focus today on word meaning, also called lexical semantics. 12/30/2021 LIN 6932 2

Outline: Comp Lexical Semantics • Intro to Lexical Semantics – Homonymy, Polysemy, Synonymy – Online resources: Word. Net • Computational Lexical Semantics – Word Sense Disambiguation • Supervised • Semi-supervised – Word Similarity • Thesaurus-based • Distributional 12/30/2021 LIN 6932 3

Preliminaries • What’s a word? – Notions we’ve used so far: Types, tokens, stems, roots, inflected forms, etc. . . – Lexeme: An entry in a lexicon consisting of a pairing of a form with a single meaning representation 12/30/2021 LIN 6932 4

Lexeme • Lexeme – a pairing of a form with a sense • Orthographic form – the way the lexeme looks on the page • Phonological form – the way the lexeme sounds • Lexicon – finite list of lexemes eaten eat Lexeme eat ate 12/30/2021 LIN 6932 5

Lexical Relations: Relationships between word meanings • • Homonymy Polysemy Synonymy Hypernymy Meronymy Antonymy 12/30/2021 LIN 6932 6

Homonymy A relation that holds between two lexemes that have the same form (phonological, orthographic or both) with unrelated meanings • Homophones right - write piece - peace • Homographs bass (fish) - bass (guitar) • Homographs (and also homophones) Bat (wooden stick-like thing) Bat (flying scary mammal thing) Bank (financial institution) Bank (riverside) 12/30/2021 LIN 6932 7

Homonymy A relation that holds between two lexemes that have the same form (phonological, orthographic or both) with unrelated meanings Example: BANK sloping mound Lexeme bank Lexeme 12/30/2021 LIN 6932 Financial institution 8

Homonymy causes problems • Spelling correction – Confusables – your vs. you’re • Speech recognition – Homophones and pure homonyms • Text-to-speech – Homographs (same orthographic form but different phonological form) • bass vs bass • Information retrieval – Homographs and pure homonyms (different meanings but same orthographic form) • QUERY: bat care 12/30/2021 LIN 6932 9

Polysemy a single lexeme has multiple related meanings While some banks furnish sperm only to married women, others are less restrictive. I withdrew the money from the bank. The bank is constructed from red brick. Biological repository bank Lexeme Financial institution 12/30/2021 LIN 6932 10

Polysemy • A single lexeme with multiple related meanings (bank the building, bank the financial institution) • Most common words have multiple meanings – The number of meanings is related to its frequency – Verbs tend more to polysemy 12/30/2021 LIN 6932 11

Specific types of polysemy: Metaphor and Metonymy • Metaphor: – Sally is cold as ice. – The relationship is on the rocks. • Metonymy – The White House announced its decision yesterday. – Shakespeare takes up 5 feet of my bookshelves. 12/30/2021 LIN 6932 12

Metaphor • Metaphors help us understand one domain of experience (typically some abstract domain) in terms of another (typically a concrete domain). Sally is a block of ice is easy to understand but Sally is a table is odd. The reason is that • we systematically and ‘habitually’ understand the domain of interpersonal relationships in terms of temperature. 12/30/2021 LIN 6932 13

Metaphor • Time understood in terms of space The end of semester is coming upon us. 12/30/2021 LIN 6932 14

Metonymy The use of one concept (domain) to refer to another concept (domain) whereby both belong to the same encompassing experiential (superordinate) domain PRODUCER FOR PRODUCT • I’ll have an Anchorsteam. AUTHOR FOR WORKS • He likes Shakespeare. PLACE FOR INSTITUTION (people who constitute that institution) • The White House had no comment. OBJECT USED FOR THE USER • The gun he hired wanted fifty grand. PART FOR WHOLE • 12/30/2021 We don’t hire skinheads. LIN 6932 15

How do we know when a word has more than one sense? • airline examples Which flights serve breakfast? Does Delta serve Philadelphia? • The “zeugma” test/effect: ? Does United serve breakfast and San Jose? • Distinct meanings of an ambiguous lexeme (homophone, homograph) can give rise to the zeugma effect - ‘oddity’ 12/30/2021 LIN 6932 16

How do we know when a word has more than one sense? The “zeugma” test/effect: ? Does Delta serve breakfast and San Jose? Coordination construction with ellipsis Does Delta serve breakfast and <does Delta serve> San Jose? 1. Syntactic constraint on the coordinands (units) in a coordination construction: they are typically of the same syntactic category: e. g. , V, VP, clause, etc. 2. Semantic constraint: If a part of the second coordinand is elided, it must be interpreted in a parallel (ideally identical) fashion with the first coordinand. 3. Hence, due to the parallelism, the choice of a reading for ‘serve’ in the first coordinand determines the reading of the understood ‘serve’ in the second elided constituent. The choice of a distinct meaning of the ambiguous lexeme ‘serve’ gives rise to a zeugma effect in the above example 12/30/2021 LIN 6932 17

How do we know when a word has more than one sense? Another example: ? John took some money out of his savings and then a vacation. The corresponding (understood) not elided coordination construction: ? John took 1 some money out of his savings and then he took 1 a vacation. 12/30/2021 LIN 6932 18

How do we know when a word has more than one sense? Yet another example: Zeugma test and VP anaphora (1) John saw a mole and so did Bill. The corresponding (understood) coordination construction (2) John saw a mole and Bill saw a mole. Due to the parallelism, the choice of a reading for ‘a mole’ in the first coordinand determines its reading in the second VP in (2), and also in the corresponding VP anaphor ‘so did’ in (1) 12/30/2021 LIN 6932 19

How do we know when a word has more than one sense? • Distinct meanings of a polysemous lexeme (homophone, homograph) do NOT give rise to zeugma effects - ‘oddity’ Pick up the glass 1, and pour it 1 into the pitcher. (example from Green 1989, p. 48) • Here, different meanings of a single polysemous lexeme ‘glass’ (container, content) are used in different clauses without oddity 12/30/2021 LIN 6932 20

Synonymy • Two lexemes are synonyms if they can be successfully substituted for each other in all situations without changing the content of a whole sentence (preservation of truth conditions) • Some words that have the same meaning in some contexts. – – – filbert youth big automobile couch soda hazelnut adolescent large car sofa pop • There are probably no exact synonyms What are some differences between “big” and “large”? 12/30/2021 LIN 6932 21

Synonymy How big is that plane? How large is that plane? A big fat apple ? A large fat apple a big brother a large brother older big lexeme Positive size large 12/30/2021 lexeme LIN 6932 22

Hyponymy A relation that hold between two lexemes where one denotes a subclass of the other vehicle car • ontology • taxonomy • object hierarchy 12/30/2021 hypernym vehicle hyponym car LIN 6932 23

Hyponymy • What is hyponymy? • Not symmetric – Example: car is a hyponym of vehicle and vehicle is a hypernym of car – Test: That is a car implies That is a vehicle (but not vice versa) • What is an ontology? – Ex: CAR#1 is an object of type car • What is a taxonomy? – Ex: car is a kind of vehicle. CAR#1 is an object of type car • What is an object hierarchy? 12/30/2021 LIN 6932 24

Resources • There are lots of lexical resources available these days… – Word lists – On-line dictionaries – Corpora • The most ambitious one is Word. Net – A database of lexical relations for English • Versions for other languages are under development 12/30/2021 LIN 6932 25

Word. Net Relations 12/30/2021 LIN 6932 26

Word. Net Hierarchies 12/30/2021 LIN 6932 27