Word Meaning IV PDF

Word Meaning IV HUL 243 9th Sep 2024 Ashwini Vaidya (HUL 243) 9th Sep 2024 1 / 23 Word relations Ambiguity, vagueness and polysemy Psychological experiments on ambiguity Ashwini Vaidya (HUL 243)...

Word Meaning IV HUL 243 9th Sep 2024 Ashwini Vaidya (HUL 243) 9th Sep 2024 1 / 23 Word relations Ambiguity, vagueness and polysemy Psychological experiments on ambiguity Ashwini Vaidya (HUL 243) 9th Sep 2024 2 / 23 Synonymy Synonyms are phonological words which have the same or very similar meaning I English appears to have several synonyms due to the influence from other languages I Below are some synonyms from English with diverse origin Anglo-Saxon brotherly world kingly French fraternal universe royal Latin regal I Usually, one of the synonyms is more in common use (e.g. royal is more common than kingly I Generally though, why would two words with exactly the same meaning survive in the language? Ashwini Vaidya (HUL 243) 9th Sep 2024 3 / 23 Origin of Synonyms 1 Dialects of the language US English fall faucet British English autumn tap We note that broadly, there are differences in terms of lexicon, pronounciation in British and American English (although they are mutually intellegible) Hence they are considered dialects of English: as are Singapore English, Australian English, Indian English etc. Within British English there are also regional varieties (London, Liverpool etc). Similar is the case with the US Ashwini Vaidya (HUL 243) 9th Sep 2024 4 / 23 1 Register: variety of a language used for a particular situation Neutral nasty man Formal obnoxious gentleman Colloquial yucky chap In the above example, the three are ostensibly synonyms, but can be used appropriately only in a given social situation Ashwini Vaidya (HUL 243) 9th Sep 2024 5 / 23 Distributional effects in Synonymy 1 Synonyms may not always occur in the same contexts: they may be collocationally restricted a big house a big sister a large house ?a large sister strong tea strong man ?powerful tea powerful man Ashwini Vaidya (HUL 243) 9th Sep 2024 6 / 23 Distributional effects in Synonymy 1 Synonyms may not always occur in the same contexts: they may be collocationally restricted a big house a big sister a large house ?a large sister strong tea strong man ?powerful tea powerful man If substitution is a test for synonymy, then it doesn’t always work It appears that words are completely interchangeable in certain environments only E.g. deep and profound can be used with sympathy but not water This seems to indicate that there are no true synonyms Ashwini Vaidya (HUL 243) 9th Sep 2024 6 / 23  I’ll go to the shop and some bread Can we have ‘true’ synonyms here? The items only appear synonymous in the similar context Ashwini Vaidya (HUL 243) 9th Sep 2024 7 / 23 Antonymy The term antonymy is used for ‘oppositeness of meaning’: words that are opposite are antonyms While languages may/may not have a true need for synonymy, antonymy is a regular and natural feature of the language However, there are different kinds of ‘oppositeness’ and we should distinguish between them Ashwini Vaidya (HUL 243) 9th Sep 2024 8 / 23 Simple antonymy This is a relation between words when the negative of one implies the positive of the other. These are also called complementary pairs or binary pairs I dead or alive (of animals) I pass or fail I hit or miss ??? My pet python is dead but luckily it’s still alive Ashwini Vaidya (HUL 243) 9th Sep 2024 9 / 23 Gradable antonyms These are opposites where the positive of one term does not imply the negative of the other I rich -poor I young - old I beautiful - ugly All these pairs have in common the fact that they may be seen in terms of degrees of the quality involved A person may be rich or very rich, and another may be richer than the other Each of these terms can only be understood in terms of being older, richer, younger than something Intermediate terms may exist: hot (warm, tepid, cool) cold Ashwini Vaidya (HUL 243) 9th Sep 2024 10 / 23 Reverses Reverse relations can be described between terms that involve movement I push - pull I come - go I ascend - descend I up - down I expand - contract I fill -empty These describe movement in one direction vs. the opposite direction; also to processes that can be reversed Ashwini Vaidya (HUL 243) 9th Sep 2024 11 / 23 Converses Converses describe a relation between two entities from alternate viewpoints I above - below I employer - employee I own - belong to Alan owns this book – This book belongs to Alan Helen is David’s employer – David is Helen’s employee These also explain paraphrases: I My office is above the library I The library is below my office Ashwini Vaidya (HUL 243) 9th Sep 2024 12 / 23 Meronymy Meronymy is a term used to describe part-whole relationship between lexical items The terms cover and page are meronyms of book We can identify this relationship using a test like X is a part of Y : A page is a part of a book Meronymy reflects some hierarchical classifications in the lexicon Ashwini Vaidya (HUL 243) 9th Sep 2024 13 / 23  Meronymic hierarchy Meronyms may vary in how necessary the part is to the whole nose is a meronym of face but collar need not be a meronym of shirt car wheel engine door window valve piston Ashwini Vaidya (HUL 243) 9th Sep 2024 14 / 23 Hyponymy and Meronymy Hyponymy is transitive: if a kestrel is a hyponym of hawk, and hawk is a hyponym of bird, then kestrel is a hyponym of bird Meronymy may or may not be transitive: nail is a meronym of finger and finger is a meronym of hand → transitive But pane- window ; window – room but pane is not a meronym of room It is possible to segment an item in countless ways- but only some of these are encoded in the language Ashwini Vaidya (HUL 243) 9th Sep 2024 15 / 23 Other relations Member- collection I ship - fleet I tree - forest I bird - flock I fish - shoal Portion -mass I drop of liquid I grain of salt I strand of hair I sheet of paper Ashwini Vaidya (HUL 243) 9th Sep 2024 16 / 23 A lexical network WordNet developed by linguists and psychologists at Princeton (Miller, Felbaum etc) Organization based on a division into 4 categories: nouns, verbs, adjectives and adverbs (no function words) Within these categories, words are hierarchically arranged English WordNet was followed by others in many languages (IndoWordNet) WordNet can be viewed as a graph whose nodes are synsets and edges are semantic relations between those synsets Ashwini Vaidya (HUL 243) 9th Sep 2024 17 / 23 Lexical The word bank when used as a noun has the following synsets in WordNet Ashwini Vaidya (HUL 243) 9th Sep 2024 18 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses I A word is a string- it represents a single linguistic element e.g. bank Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses I A word is a string- it represents a single linguistic element e.g. bank I A synset is a set of all senses expressing the same concept e.g.{depository financial institution, bank, banking concern, banking company} belong to the same synset Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses I A word is a string- it represents a single linguistic element e.g. bank I A synset is a set of all senses expressing the same concept e.g.{depository financial institution, bank, banking concern, banking company} belong to the same synset I Given the synset and the word, a particular sense can be uniquely identified e.g. bank.n.1 Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses I A word is a string- it represents a single linguistic element e.g. bank I A synset is a set of all senses expressing the same concept e.g.{depository financial institution, bank, banking concern, banking company} belong to the same synset I Given the synset and the word, a particular sense can be uniquely identified e.g. bank.n.1 Synsets are the primary nodes used to build the semantic network of WordNet Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses I A word is a string- it represents a single linguistic element e.g. bank I A synset is a set of all senses expressing the same concept e.g.{depository financial institution, bank, banking concern, banking company} belong to the same synset I Given the synset and the word, a particular sense can be uniquely identified e.g. bank.n.1 Synsets are the primary nodes used to build the semantic network of WordNet In general the most frequently occurring synset for a word is listed first Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23  E.g sloping land is a synonym for bank in the riverbank sense and so is a depository financial institution for bank WordNet makes distinctions between words, synsets and senses I A word is a string- it represents a single linguistic element e.g. bank I A synset is a set of all senses expressing the same concept e.g.{depository financial institution, bank, banking concern, banking company} belong to the same synset I Given the synset and the word, a particular sense can be uniquely identified e.g. bank.n.1 Synsets are the primary nodes used to build the semantic network of WordNet In general the most frequently occurring synset for a word is listed first WordNet senses are famously fine-grained and detailed Ashwini Vaidya (HUL 243) 9th Sep 2024 19 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank I Hyponymy: denotes a subordinate relation e.g. table is a hyponym of furniture Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank I Hyponymy: denotes a subordinate relation e.g. table is a hyponym of furniture I Meronymy: denotes a part-of/has-member relationship e.g. leg is a meronym of table Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank I Hyponymy: denotes a subordinate relation e.g. table is a hyponym of furniture I Meronymy: denotes a part-of/has-member relationship e.g. leg is a meronym of table I Holonymy: denotes a member-of relationship e.g. copilot is a holonym of crew Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank I Hyponymy: denotes a subordinate relation e.g. table is a hyponym of furniture I Meronymy: denotes a part-of/has-member relationship e.g. leg is a meronym of table I Holonymy: denotes a member-of relationship e.g. copilot is a holonym of crew For nouns WordNet has is-a hierarchies (inheritance hierarchies) e.g. hypernymy and hyponymy Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank I Hyponymy: denotes a subordinate relation e.g. table is a hyponym of furniture I Meronymy: denotes a part-of/has-member relationship e.g. leg is a meronym of table I Holonymy: denotes a member-of relationship e.g. copilot is a holonym of crew For nouns WordNet has is-a hierarchies (inheritance hierarchies) e.g. hypernymy and hyponymy However, for verbs these are entailment relations e.g crawl entails move forward Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical relations Synsets are connected to each other via various lexical relations. These are relations between senses not the words themselves I Hypernymy: denotes a superordinate relation e.g. slope is a hypernym of bank I Hyponymy: denotes a subordinate relation e.g. table is a hyponym of furniture I Meronymy: denotes a part-of/has-member relationship e.g. leg is a meronym of table I Holonymy: denotes a member-of relationship e.g. copilot is a holonym of crew For nouns WordNet has is-a hierarchies (inheritance hierarchies) e.g. hypernymy and hyponymy However, for verbs these are entailment relations e.g crawl entails move forward Adjectives and adverbs are not arranged in is-a hierarchies (other relations like antonymy e.g. good and bad) Ashwini Vaidya (HUL 243) 9th Sep 2024 20 / 23 Lexical Knowledge Bases Resources like WordNet are examples of Lexical Knowledge Bases (LKB), used to compute various lexical relationships between words While WordNet (including its multilingual versions) is one of the oldest, and most well-known LKB, others have also been used:- I Digital versions of thesauri like Roget’s Thesaurus for English Ashwini Vaidya (HUL 243) 9th Sep 2024 21 / 23 Lexical Knowledge Bases Resources like WordNet are examples of Lexical Knowledge Bases (LKB), used to compute various lexical relationships between words While WordNet (including its multilingual versions) is one of the oldest, and most well-known LKB, others have also been used:- I Digital versions of thesauri like Roget’s Thesaurus for English I Wikitionary, a web-based collaborative dictionary Ashwini Vaidya (HUL 243) 9th Sep 2024 21 / 23 Lexical Knowledge Bases Resources like WordNet are examples of Lexical Knowledge Bases (LKB), used to compute various lexical relationships between words While WordNet (including its multilingual versions) is one of the oldest, and most well-known LKB, others have also been used:- I Digital versions of thesauri like Roget’s Thesaurus for English I Wikitionary, a web-based collaborative dictionary I Wikipedia an ‘indirect’ LKB whose nodes are Wiki pages and hyperlinks are relations between different pages Ashwini Vaidya (HUL 243) 9th Sep 2024 21 / 23 Lexical Knowledge Bases Resources like WordNet are examples of Lexical Knowledge Bases (LKB), used to compute various lexical relationships between words While WordNet (including its multilingual versions) is one of the oldest, and most well-known LKB, others have also been used:- I Digital versions of thesauri like Roget’s Thesaurus for English I Wikitionary, a web-based collaborative dictionary I Wikipedia an ‘indirect’ LKB whose nodes are Wiki pages and hyperlinks are relations between different pages I BabelNet interlinks WordNet and Wikipedia and is available in many languages Ashwini Vaidya (HUL 243) 9th Sep 2024 21 / 23 Similarity measures Semantic similarity usually refers to the idea that two concepts may be substituted in the same context e.g. cute and pretty Ashwini Vaidya (HUL 243) 9th Sep 2024 22 / 23 Similarity measures Semantic similarity usually refers to the idea that two concepts may be substituted in the same context e.g. cute and pretty Similarity is often confused with relatedness, where two senses may be correlated in some way- but cannot be substituted e.g. beautiful and appeal Ashwini Vaidya (HUL 243) 9th Sep 2024 22 / 23 Similarity measures Semantic similarity usually refers to the idea that two concepts may be substituted in the same context e.g. cute and pretty Similarity is often confused with relatedness, where two senses may be correlated in some way- but cannot be substituted e.g. beautiful and appeal Semantic similarity can be measured via the use of LKBs like WordNet Ashwini Vaidya (HUL 243) 9th Sep 2024 22 / 23 Similarity measures Semantic similarity usually refers to the idea that two concepts may be substituted in the same context e.g. cute and pretty Similarity is often confused with relatedness, where two senses may be correlated in some way- but cannot be substituted e.g. beautiful and appeal Semantic similarity can be measured via the use of LKBs like WordNet The taxonomic structure of the LKB is exploited in order to calculate the degree of similarity between two concepts Ashwini Vaidya (HUL 243) 9th Sep 2024 22 / 23 Similarity measures Semantic similarity usually refers to the idea that two concepts may be substituted in the same context e.g. cute and pretty Similarity is often confused with relatedness, where two senses may be correlated in some way- but cannot be substituted e.g. beautiful and appeal Semantic similarity can be measured via the use of LKBs like WordNet The taxonomic structure of the LKB is exploited in order to calculate the degree of similarity between two concepts When we use a structured LKB like WordNet to compute similarity, it is a knowledge-based approach to semantic similarity Ashwini Vaidya (HUL 243) 9th Sep 2024 22 / 23 References Semantics by John Saeed An overview of word and sense similarity by Roberto Navigli and Federico Martelli, NLE (2019) Vol 25, 693-714 Five papers on WordNet, Miller et al (1993) Matthew Traxler, Introduction to Psycholinguistics (2011) Wiley Ashwini Vaidya (HUL 243) 9th Sep 2024 23 / 23  Dual route/Situations HUL 243 23rd Sep 2024 Ashwini Vaidya (HUL 243) 23rd Sep 2024 1 / 19 Two approaches differ based on storage Ashwini Vaidya (HUL 243) 23rd Sep 2024 2 / 19 Whole-word lexicon In the whole-word lexicon approach, there is an idea that storage is large, unlimited Morphemes are not the basic units that are stored, but larger chunks (words themselves or even larger chunks) are also possible The whole word approach has far less restrictiveness and much more redundancy Emphasis on system-external explanations, such as language use Ashwini Vaidya (HUL 243) 23rd Sep 2024 3 / 19 Past tense aquisition in children Hoeffner et al, (1993) re-examined the child language acquisition data on past tense inflections Children also produced unmarked verb forms such as I play-∅ with the baby rabbits yesterday The past tense was not produced in an obligatory context Therefore, the over-regularization was just a part of a more general change Children went from unmarked play to played or holded Eventually, they switched to hold-holded but this was based on a gradual change (not a sudden switch) Ashwini Vaidya (HUL 243) 23rd Sep 2024 4 / 19 Emphasis on experience System-external motivation for learning process Based on an idea of maximal storage, storage is cheap! Joan Bybee’s ‘The Emergent Lexicon’ “I am going to argue that if such a lexicon does exist it is because it is emergent from the storage of linguistic experience, which is of a very different nature than the traditional conception of a lexicon would suggest. ” Ashwini Vaidya (HUL 243) 23rd Sep 2024 5 / 19 Storage/Memory We store a very large number of details: visual, auditory, olfactory etc. In a similar way, we are sensitive to linguistic experience from an early age Fairly powerful pattern recognition capability, even with very little exposure Saffran, Aslin and Newport (1996) Infants exposed to only two minutes of an artificial language were able to infer word boundaries in test scenarios, i.e. able to distinguish ‘words’ they had heard- and ‘non-words’ they hadn’t during training. Ashwini Vaidya (HUL 243) 23rd Sep 2024 6 / 19 Storage/Memory Not only can we extract patterns, but we can also recall them (after long periods of time) E.g. recalling voices, words, sentences, phrases even after brief exposure Standing (1973) Demonstrated the power of visual memory:- participants saw 10,000 pictures (people on vacation) over a course of a week, and were able to recall quite reliably whether they had seen a given picture before or not. When we have unlimited storage, we don’t need to minimize the size of the lexicon We store the monolithic surface forms (without attending to their individual structure) e.g yellow brick road or ferris wheel Ashwini Vaidya (HUL 243) 23rd Sep 2024 7 / 19 Patterns Does this mean there is no room for generalization/productivity/creativity? Analogy: form-meaning correspondences are used to produce some generalization If certain patterns occur often enough during linguistic experience, we can predict how other forms might behave E.g wugs → bugs, mugs, rugs, hugs Ashwini Vaidya (HUL 243) 23rd Sep 2024 8 / 19 Over-regularization How to account for phenomena like over-regularization ? Analogy and experience together can explain the child language data of holded and heared 1 Initially, no past tense forms are produced, even in obligatory contexts 2 Then, via analogistic pressure, there is a form-meaning correspondence between regular past -ed and irregular verbs hold, hear 3 Gradual increase in input over time competes/does not agree with the analogical forms 4 Finally, there is recovery from over-generalization More than one cue for learning is possible, importance of experience Ashwini Vaidya (HUL 243) 23rd Sep 2024 9 / 19 Strict word-form lexicon If a notion of a word is the largest possible ‘chunk’ of meaning, then it works with languages like English, German which have fewer affixes However – it still does not explain how speakers learn agglutinative language structures, where each word consists of many morphemes (1) oku-ya-ma-yabil-ir-im read-POT-NEG-POT-AOR-1SG ‘I might not be able to read’ There also exist issues with assuming that words have no internal structure (there seems evidence that they do) Further a ’whole-word’ lexicon would be a lot less elegant and modular than a morpheme based lexicon Ashwini Vaidya (HUL 243) 23rd Sep 2024 10 / 19  Figure: Schema of the dual-route model, Hay (2001) Ashwini Vaidya (HUL 243) 23rd Sep 2024 11 / 19 Moderate word-form lexicon Insane is more common than the root sane, more frequently used Hence insane has better memory strength and is retrieved via direct route On the other hand, if insanely needs to be retrieved, it is also less frequent than insane– does it get decomposed? Entirely possible that in this model, some partial decomposition takes place e.g. insane+ ly rather than in+sane+ly The ‘word form’ that is memorized does not need to correspond to a minimal, analytical form Ashwini Vaidya (HUL 243) 23rd Sep 2024 12 / 19 Situations Ashwini Vaidya (HUL 243) 23rd Sep 2024 13 / 19 Jack and Jill went up the hill, To fetch a pail of water. Jack fell down and broke his crown, And Jill came tumbling after Ashwini Vaidya (HUL 243) 23rd Sep 2024 14 / 19  When we hear this, we get the intuition that there are some discrete events What are these events ? Which linguistic structures express them ? In the end, these are connected into a coherent whole How are different kinds of events encoded in language? Ashwini Vaidya (HUL 243) 23rd Sep 2024 15 / 19 Situation types Verbs can be described in terms of situation types Situation type describes the events encoded in the semantics of a language This is most commonly encoded in verbs Ashwini Vaidya (HUL 243) 23rd Sep 2024 16 / 19 Situation types: states For example, some verbs describe states:- I Robert loves pizza I Mary knows the way to the LHC In such examples, we don’t know much about the internal structure of the state: it just holds for some point The beginning or end of this situation is not known (e.g. Even if I have Mary loved pizza – we don’t know when she stopped doing so Verbs like be, have, remain, know, love are stative verbs Ashwini Vaidya (HUL 243) 23rd Sep 2024 17 / 19  Model of the lexicon HUL 243: Language and Communication 23rd September 2024 Model of the lexicon 1 / 26 Lexicon I Words in the language user’s vocabulary are stored in the mental lexicon I Words must be listed in the lexicon for them to be a part of it I What is the actual form of this listing? – Are words stored whole e.g. readable, washable, reads – Are words stored as morphemes e.g. read, wash, -able lexicon whole-words morphemes readable, reads, washable -able, read, wash Model of the lexicon 2 / 26 Morpheme A Morpheme is the smallest linguistic unit with a grammatical function (Aronoff) I Let’s look at the word reconsideration I It consists of three parts re- + consider + -ation I consider is the stem i.e. a base unit to which other morphemes may attach I re- and -ation are both affixes; re- is a prefix and -ation is a suffix I reconsideration consists of 3 morphemes Model of the lexicon 3 / 26 Lexical access I Lexical access is the process of looking up a word in the lexicon I If both morphemes and whole words are stored in the lexicon:- I Two potential ways to look up a given word- I Decomposition route i.e. access words by breaking them into morphemes I Direct route i.e. retrieving words without decomposition I E.g. a word like insane is retrieved as in- + sane or insane Model of the lexicon 4 / 26 I Dual route model suggests both routes are in use: decomposition and direct access I The winner is whichever method is faster in accessing the information Figure: Schema of the dual-route model, Hay (2001) Model of the lexicon 5 / 26 Morpheme A Morpheme is the smallest linguistic unit with a grammatical function (Aronoff) I Let’s look at the word reconsideration I It consists of three parts re- + consider + -ation I consider is the stem i.e. a base unit to which other morphemes may attach I re- and -ation are both affixes; re- is a prefix and -ation is a suffix I reconsideration consists of 3 morphemes Model of the lexicon 6 / 26 Children’s acquisition of words I First-language learners speak a language without explicit instruction by age 3 I Somehow, children distil out the essential rules of the language based on the finite set of utterances they have heard I How do children make this leap? I Children generalize from the set of of linguistic structures they hear Model of the lexicon 7 / 26 Children’s acquisition of words I First-language learners speak a language without explicit instruction by age 3 I Somehow, children distil out the essential rules of the language based on the finite set of utterances they have heard I How do children make this leap? I Children generalize from the set of of linguistic structures they hear My teacher holded the baby rabbits and we patted them I finded Renee Hey I heared a sound Model of the lexicon 7 / 26 Children’s acquisition of words I Children learn something like ‘to form the past tense, add -ed’ I When encountering find, hear, hold → found, heard, held, they over-regularize I Shows that an internal awareness of the past tense formation rule I Note that children often use both forms held-*holded, went-*goed for some time without realizing it I At some time, the switch flips– when they realize that held, went are past (like the -ed forms) Model of the lexicon 8 / 26 Children’s acquisition of words I This developmental pattern seems to suggest the following:- I Regular complex words like show-showed are likely to be listed in the mental lexicon as combinations of roots and affixes I Perhaps, the idiosyncratic or irregular forms are stored (as these are a minority) I Words are formed via computation, structure building I This view supports the idea of the lexicon as morpheme-based Model of the lexicon 9 / 26 Morpheme-based lexicon I Children eventually learn to build (generate) the rules for existing forms I They also learn to constrain the rules to not generate ungrammatical (non-existent) forms I When children learn the elements from an inventory of word-forms they create via analysis I A morpheme-based lexicon places emphasis on assembling the pieces of words I Another way to characterize this would be that it is heavy on computation I Advantage of such a view: elegance/economy of representation Model of the lexicon 10 / 26 Economy I Memorizing entire words leads to redundancy:- I Below is the paradigm for Greek lexeme filos ‘friend’ I Total of 7 morphemes: fíl-, -os, -i, -on, -u, -us, -o vs. 6 individual word forms singular plural nominative fílos fíli accusative fílo fílus genitive fílu fílon I But filos is part of an inflectional paradigm that includes kosmos ‘world’, fovos ‘fear’, gamos ‘marriage’, skilos ‘dog’ I 30 word-forms vs. 11 morphemes Model of the lexicon 11 / 26 Lexical economy I Even derivational forms can occur in families, read and write are related in a parallel fashion read reader readable write writer writable I Morpho-phonological rules of allomorphy are also economical I E.g. consonant devoicing rule [-z] → [-s] before voiceless sounds can be found in plurals and past tense: I cat → cat[s]; dog → dog[z] I work → worke[t] ; arrive → arrive[d] Model of the lexicon 12 / 26 Agglutinative languages I Agglutinating languages have a high ratio of morphemes to word I Each word form can occur with many possible affixes I Impossible to memorize that many forms! Turkish (1) oku-ya-ma-yabil-ir-im read-POT-NEG-POT-AOR-1SG ‘I might not be able to read’ Model of the lexicon 13 / 26 Modularity I At the phonological level, words consist of individual phonemes I Elements at phonology combine (‘assembled’) and feed into morphology I At the morphological level, morphmes combine to form new words I In syntax, these words combine to form phrases, then sentences Model of the lexicon 14 / 26 I Computation is cheap– storage is expensive! I Generate morphological rules from morphemes, phonological rules from phonemes I No storage in lexicon, unless absolutely needed! I Lexicon is a space that only has the following:- I Monomorphemic forms and affixes I Irregular forms (hold-held, find-found ) I Fossilized forms (ox-oxen, mouse-mice) I Semantic idiosyncrasy (suffix -ee attaching to animate nouns) I Lexicon is static, the grammar has the moving parts Model of the lexicon 15 / 26 Challenges for morpheme-based approach Irregularity I Irregularity in paradigms is very common I French paradigm for verb parler ‘speak’ je parle I speak nous parlons we speak tu parles You speak vous parlez you(pl) speak il/elle parle He/she speaks ils/elles parlent they speak I Irregular aller ‘go’ + many irregular forms of verbs ! je vais I go nous allons We go tu vas You go vous allez You(pl) go il/elle va He/She goes ils/elles vont They go I If irregular forms are stored, economy ↓ Model of the lexicon 16 / 26 Challenges for morpheme-based approach Irregularity I English suppletive forms:- I go → went must be stored (go+ past) I bad → worse must be stored (bad+ comparative) I Not a challenge per se, but does imply increased use of lexical storage Model of the lexicon 17 / 26 Challenges for morpheme-based approach Predictability of meaning I Predictability of meaning I Composition of forms should yield compositional meaning, but not always! I The items in the second column are mono-morphemic, must be stored separately X + -er X ? + -er painter cobbler lecturer butcher influencer bursar I Other meanings may not be predictable e.g. mailer ‘a pamphlet sent by post’, grinder ‘machine used for grinding’ Model of the lexicon 18 / 26 Challenges for a morpheme based approach Non-concatenative morphology I Morphological process not due to affixation I Zero expression e.g. fish (sg) → fish-∅ (pl) I Conversion e.g. repeat ( V) → repeat (N) I Base/stem modification e.g. sing → sang I Reduplication e.g. in Hindi dheere dheere Model of the lexicon 19 / 26 Non-concatenative morphology I For many of these cases, solutions have been proposed in the literature I E.g. sing → sang have been explained via a phonological rule applying in the environment of a null (∅) past-tense morpheme I Including zero morphemes implies that we still have a economical lexion– I...but also different lexical entries:- ∅ had no form, but meaning Model of the lexicon 20 / 26 I The morpheme-based approach is represented by a number of morphological theories, each of these may differ in their details I But in order to follow the ‘maximum computation, minimum storage’ maxim, the system can become complex Tiv Root Imperative Gloss kimbi kìmbí pay (Niger-Congo, Nigeria) de dé leave gba gbá fall va vá come I The imperative meaning is contained within the rule of tone assignment (but not to a particular form) Model of the lexicon 21 / 26 Two approaches differ based on storage Model of the lexicon 22 / 26 I Morpheme-based approaches minimize storage, maximize computation I Whole-word based approaches maximize storage, minimize computation Model of the lexicon 23 / 26 References I Haspelmath and Sims, Ch 4 I Steven Pinker ‘Why the child holded the baby rabbits: a case study in language acquisition’ from ‘An invitation to Cognitive Science’ eds. Gleitman, Osherson, Liberman Model of the lexicon 24 / 26  Situations HUL 243 3rd Oct 2024 Ashwini Vaidya (HUL 243) 3rd Oct 2024 1 / 12  Situation types are used to delineate the verbs into different categories We have already examined statives: we look at achievements, accomplishments and activity Ashwini Vaidya (HUL 243) 3rd Oct 2024 2 / 12  Adverbials such as for five minutes show a strong affinity for events with duration They are acceptable with activity verbs, not acceptable with accomplishment verbs Other such examples are : I Sam ran/pushed the cart for half an hour (activity) I Sam drew a circle #for half an hour (accomplishment) Ashwini Vaidya (HUL 243) 3rd Oct 2024 3 / 12  Activity predicates generally do not specify an end point/terminus, hence it’s ok for them to continue over a period of time, e.g. for half an hour or for five minutes In contrast, accomplishments describe situations which consist of successive phases that are not the same as the whole (e.g. if you ’freeze’ a point while drawing a circle, it’s not the same as the others) Accomplishments would describe events that proceed towards their endpoint I Activity: run, walk, swim, push a cart, drive a car I Accomplishment: paint a picture, grow up, deliver a sermon Ashwini Vaidya (HUL 243) 3rd Oct 2024 4 / 12  Interestingly, if we modify accomplishments with temporal in -adverbials then it’s alright This is a I Sam ran/pushed the cart #in half an hour (activity) I Sam drew a circle/ran a mile in half an hour (accomplishment) This is because unlike for- this will modify the entire event, Ashwini Vaidya (HUL 243) 3rd Oct 2024 5 / 12  Classifying Activity vs Accomplishment Ashwini Vaidya (HUL 243) 3rd Oct 2024 6 / 12 Compatibility with progressive What activity and accomplishment predicates have in common is a compatibility with the progressive aspect I John was running (activity) I John was drawing a circle (accomplishment) What does this mean? It means that these are situations that continue in time in different phases However, note the the following is not possible with the progressive I #Bill was noticing that Mary had a new hat Ashwini Vaidya (HUL 243) 3rd Oct 2024 7 / 12 Achievements Achievements are events that occur at a single moment An achievement would generally result in a new state, instantaneously E.g. find a flower, reach the top, win a race Such verbs are not compatible with the progressive Ashwini Vaidya (HUL 243) 3rd Oct 2024 8 / 12 Ashwini Vaidya (HUL 243) 3rd Oct 2024 9 / 12 Tests for Achievements Occurrence at a single moment: modifiable by adverbials that refer to specific moments in time I The detective found the victim’s body at 4 PM (achievement) I # John drew a circle/built a house at 4 PM (accomplishment) I # John ran at 4 PM (activity) Ashwini Vaidya (HUL 243) 3rd Oct 2024 10 / 12  Activity, Accomplishment, Achievement Ashwini Vaidya (HUL 243) 3rd Oct 2024 11 / 12 Semelfactive Semelfactives are instantaneous events without an end point Examples are cough, knock, sneeze, blink Such events involve such instantaneous action that it is virtually within no time It is possible to combine semelfactives with an for-adverbial for an iterative interpretation Ashwini Vaidya (HUL 243) 3rd Oct 2024 12 / 12  Situations- II HUL 243 14th Oct 2024 Ashwini Vaidya (HUL 243) 14th Oct 2024 1 / 21 Recap Discussed situation types: achievement, accomplishment and activity Situation types expressed by verbs How these are distinguished: use of progressive, adverbial for-expressions Ashwini Vaidya (HUL 243) 14th Oct 2024 2 / 21  Tense is the linguistic device to express time relations English has three basic tenses:- I She saw me (past) I She will see me (future) I She sees me (present) Generally, the reference point for the tense system is act of speaking Ashwini Vaidya (HUL 243) 14th Oct 2024 3 / 21  The reference to past and future is taken at the time of speaking Ashwini Vaidya (HUL 243) 14th Oct 2024 4 / 21  Note that tense can be expressed via a bound morpheme on the verb It may also be expressed periphrastically i.e. via a phrase I She played tennis (bound) I She will play tennis (periphrastic) I She plays tennis (bound) Ashwini Vaidya (HUL 243) 14th Oct 2024 5 / 21  In Chinese, there are no bound morphemes for coveying tense- this can be done only periphrastically It is possible to have sentences without periphrastic tense markers that convey time (1) Ta dai ni qu nail ? ta take you go where ? Where did he take you ? (2) Zhangsan dapuo yi-ge huaping Zhangsan break one-cl vase Zhangsan broke a vase Ashwini Vaidya (HUL 243) 14th Oct 2024 6 / 21  Chinese poetry exploits this ‘timeless’ quality in its poetry A direct translation has no verbs in 4 out of the 5 lines Autumn Thoughts by Ma Zhiyuan – literal translation Dried vines, old tree, evening crows Small bridge, running water, people homes Old road, west wind, scrawny horse Evening sun west sets Guts-torn man at sky’s end Ashwini Vaidya (HUL 243) 14th Oct 2024 7 / 21  Chinese poetry exploits this ‘timeless’ quality in its poetry A direct translation has no verbs in 4 out of the 5 lines Autumn Thoughts by Ma Zhiyuan – literal translation Dried vines, old tree, evening crows Small bridge, running water, people homes Old road, west wind, scrawny horse Evening sun west sets Guts-torn man at sky’s end A more appealing translation.. Withered wisteria, old tree, darkling crows Little bridge over flowing water by someone’s house Emaciated horse on an ancient road in the western wind Evening sun setting in the west Broken-hearted man on the horizon. Ashwini Vaidya (HUL 243) 14th Oct 2024 7 / 21 More on Tense Tense does not generally tell us about the duration of time. E.g. the following sentences are in the present tense- but are really intervals including the present moment– the length of these intervals varies I He is hungry I Smoking kills I That is amazing! I Brown defends the freedom of the press in his novel Ashwini Vaidya (HUL 243) 14th Oct 2024 8 / 21 More on tense The past tense refers to a time in the past, without telling us about its connection to the present e.g. Did you finish ? If we wish to refer to a past time periphrastically, then we can use adverbs like yesterday and last year Ashwini Vaidya (HUL 243) 14th Oct 2024 9 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ I I have written, 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ I I have written, I I had written1 , 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ I I have written, I I had written1 , I I will have written 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ I I have written, I I had written1 , I I will have written e.g. Have you finished ? implies whether the action (as of this current moment) is completed or not ? 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ I I have written, I I had written1 , I I will have written e.g. Have you finished ? implies whether the action (as of this current moment) is completed or not ? Other examples: Have you read Shakespeare’s sonnets ? - implies that as of this current moment, you either read the sonnets or not 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Perfect In contrast to past tense, the perfect refers to some time in the past, and connects it to the present Formed by adding ‘have/had’ I I have written, I I had written1 , I I will have written e.g. Have you finished ? implies whether the action (as of this current moment) is completed or not ? Other examples: Have you read Shakespeare’s sonnets ? - implies that as of this current moment, you either read the sonnets or not On the other hand, if you were referring to a literature course taken by your friend, you may ask –in the past tense – Did you read Shakespeare’s sonnets ? 1 Also known as pluperfect Ashwini Vaidya (HUL 243) 14th Oct 2024 10 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening I I will be listening Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening I I will be listening E.g. He was writing when I entered means that his writing has started but not completed at the moment I entered Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening I I will be listening E.g. He was writing when I entered means that his writing has started but not completed at the moment I entered This form used to show that an action is incomplete rather than complete (I am finishing the manuscript) Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening I I will be listening E.g. He was writing when I entered means that his writing has started but not completed at the moment I entered This form used to show that an action is incomplete rather than complete (I am finishing the manuscript) or that an act is habitual rather than momentary (I was running on weekday evenings) Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening I I will be listening E.g. He was writing when I entered means that his writing has started but not completed at the moment I entered This form used to show that an action is incomplete rather than complete (I am finishing the manuscript) or that an act is habitual rather than momentary (I was running on weekday evenings) or intentions/plans in the immediate future (I am catching the midnight train tonight) Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Complex tenses: Progressive The progressive shows an action or state with a time frame that is continuous or expanded Formed by adding be + -ing I I am listening, I I was listening I I will be listening E.g. He was writing when I entered means that his writing has started but not completed at the moment I entered This form used to show that an action is incomplete rather than complete (I am finishing the manuscript) or that an act is habitual rather than momentary (I was running on weekday evenings) or intentions/plans in the immediate future (I am catching the midnight train tonight) Note that it’s possible to combine both complex tenses: I have been waiting since afternoon is present perfect progressive Ashwini Vaidya (HUL 243) 14th Oct 2024 11 / 21 Difference between the two sentences? Ralph was building a fire escape last week Ralph built a fire escape last week In the first case, the action is not yet completed - and we don’t know when it will: this is also known as imperfective aspect Ashwini Vaidya (HUL 243) 14th Oct 2024 12 / 21 Difference between the two sentences? Ralph was building a fire escape last week past progressive Ralph built a fire escape last week past In the first case, the action is not yet completed - and we don’t know when it will: this is also known as imperfective aspect In the second case, the action is complete- and the sentence is known to have perfective aspect Ashwini Vaidya (HUL 243) 14th Oct 2024 12 / 21 Difference between the two sentences? Ralph was building a fire escape last week past progressive Ralph built a fire escape last week past In the first case, the action is not yet completed - and we don’t know when it will: this is also known as imperfective aspect In the second case, the action is complete- and the sentence is known to have perfective aspect Aspect is not explicitly marked on the verb in English -it is not grammaticalized, but this is the case with languages like Spanish and Russian Ashwini Vaidya (HUL 243) 14th Oct 2024 12 / 21  Examine the story below and answer the question Those people, not one of them realized I was not human. They looked at me and they pretended I was someone called Sister Mary. Maybe they really thought I was Sister Mary. In the story above, was I Sister Mary? Ashwini Vaidya (HUL 243) 14th Oct 2024 13 / 21  Examine the story below and answer the question Those people, not one of them realized I was not human. They looked at me and they pretended I was someone called Sister Mary. Maybe they really thought I was Sister Mary. In the story above, was I Sister Mary? Ashwini Vaidya (HUL 243) 14th Oct 2024 13 / 21  Verbs such as realize are factive : explaining a true event that is realized in the complement Of course, the negation of these implies that the event is not true (occurrence of both these verbs implies that The verb thought is relatively neutral-: replacing the sentence with Maybe they really believed I was Sister Mary yielded a much better response from ChatGPT Ashwini Vaidya (HUL 243) 14th Oct 2024 14 / 21 Evidentiality How does a speaker communicate an attitude towards the source of information? E.g. a simple sentence like She was rich can be compared with the following:- I I saw that she was rich I I read that she was rich I She was rich, so they say I I’m told she was rich I Apparently she was rich I She was rich, it seems I Allegedly she was rich Such qualifications allow a speaker to say whether the statement relies on first hand knowledge, was acquired from another source or whether such a source was reliable Ashwini Vaidya (HUL 243) 14th Oct 2024 15 / 21 Evidentiality Some languages mark this information using a morpheme For example, Makah (an indigenous language of N. America) will use the following suffixes: I direct experience: zero marking I inference from physical evidence: -pi:t I auditory source: - ’qadi I quotative: - wa:t Ashwini Vaidya (HUL 243) 14th Oct 2024 16 / 21  The morpheme marking of evidentiality shows that in some languages, this must be obligatorily expressed, but in others it is optional. E.g. the Jaqi languages spoken in Peru, Bolivia and Chile, accuracy on the part of the speaker is important with respect to their public reputation Hence, they must use the evidential marker obligatorily in their speech Ashwini Vaidya (HUL 243) 14th Oct 2024 17 / 21 Modality Modality is a cover term for devices which allow speakers to express degrees of commitment to or belief in a proposition Unlike evidentiality which focuses on the source of information– modality marks the degree of confidence in a statement Noah has gone to the airport (S) Modal systems allow speakers to show strong or weak commitment to the factuality of statements In the examples below, this is done via an adjective or adverb of modality I It is certain that S I It is probable that S I It is likely that S I It is possible that S Ashwini Vaidya (HUL 243) 14th Oct 2024 18 / 21 Modal auxiliaries She has left by now I She must have left by now I She might have left by now I She could have left by now I She needn’t have left by now I She couldn’t have left by now These types of verbs are modal verbs All of these signal degrees of knowledge that the speaker has : express epistemic modality Ashwini Vaidya (HUL 243) 14th Oct 2024 19 / 21  It is also possible to use modal verbs to signal obligation, responsibility and permission: these are deontic modals You can drive this car :- you are able to drive this car OR You have my permission to drive this car Modals of obligation (in decreasing order of strength) I You must drive this car I You should drive this car I You need to drive this car I You ought to drive this car Modals of permission (in decreasing order of strength) I You can drive this car I You could drive this car I You might drive this car Ashwini Vaidya (HUL 243) 14th Oct 2024 20 / 21  Legal documents in particular contain examples of deontic modality: e.g. lessee shall pay lessor annual base rent in the amount of one hundred dollars etc. The modal auxiliary shall tells us that the lessee is obliged to pay something to the lessor To automatically process legal documents, and summarize the obligations, entitlements, prohibitions of a contracting party, understanding deontic modals is important Ashwini Vaidya (HUL 243) 14th Oct 2024 21 / 21  Participants Ashwini Vaidya 17th Oct 2024 Ashwini Vaidya Participants 1 / 17 We looked at aspects of sentence level semantics in the last few classes How do speakers characterize situations and express degrees of commitment How do we portray the roles of the entities involved? Ashwini Vaidya Participants 2 / 17 Chomsky (1965) proposed to divide up all verbs into subcategories based on their grammatical frames. I put [NP PP] (put the shoes under the desk) I eat [NP] (ate lunch, ate at noon) I contribute [NP PP] (contribute some money; contribute to the Red Cross) Here NP stands for Noun Phrase and PP for Verb Phrase Ashwini Vaidya Participants 3 / 17  This idea of frames based only on grammar is not useful: Same frame- V NP PP, but different content The frame can be invariable, but the content can be different 1 She drained the cash from the account 2 She drained the account of cash Different frame – same content 1 She asked me what time it was [V SCOMP] 2 She asked me the time [V NP] Ashwini Vaidya Participants 4 / 17 Evidence for a covert category Compare the two sentences: 1 John ruined the table 2 John built the table What did John do to the table? Ashwini Vaidya Participants 5 / 17 Evidence for a covert category Compare the two sentences: 1 John ruined the table 2 John built the table What did John do to the table? However, the response is felicitous in (1) not (2) Ashwini Vaidya Participants 5 / 17 Evidence for a covert category Compare the two sentences: 1 John ruined the table 2 John built the table What did John do to the table? However, the response is felicitous in (1) not (2) One type of syntactic object existed before the event took place (ruin) and the other after the event (build) Ashwini Vaidya Participants 5 / 17 Evidence for a covert category Compare the two sentences: 1 John ruined the table 2 John built the table What did John do to the table? However, the response is felicitous in (1) not (2) One type of syntactic object existed before the event took place (ruin) and the other after the event (build) In both sentences, table is a grammatical object but appears to have slightly different meanings Ashwini Vaidya Participants 5 / 17 All the below are subjects, but are also agents, recipients, patients and possessors 1 He hit the ball 2 He received a blow 3 He received a gift 4 He has black hair Ashwini Vaidya Participants 6 / 17 All the below are subjects, but are also agents, recipients, patients and possessors 1 He hit the ball 2 He received a blow 3 He received a gift 4 He has black hair There is a link between these semantically-typed verb arguments and their respective predicates Ashwini Vaidya Participants 6 / 17 All the below are subjects, but are also agents, recipients, patients and possessors 1 He hit the ball 2 He received a blow 3 He received a gift 4 He has black hair There is a link between these semantically-typed verb arguments and their respective predicates Fillmore’s paper ‘The Case for Case’ termed these as case relations Ashwini Vaidya Participants 6 / 17 All the below are subjects, but are also agents, recipients, patients and possessors 1 He hit the ball 2 He received a blow 3 He received a gift 4 He has black hair There is a link between these semantically-typed verb arguments and their respective predicates Fillmore’s paper ‘The Case for Case’ termed these as case relations Case relations (in Fillmore’s sense) do not need to be realized morphologically Ashwini Vaidya Participants 6 / 17 All the below are subjects, but are also agents, recipients, patients and possessors 1 He hit the ball 2 He received a blow 3 He received a gift 4 He has black hair There is a link between these semantically-typed verb arguments and their respective predicates Fillmore’s paper ‘The Case for Case’ termed these as case relations Case relations (in Fillmore’s sense) do not need to be realized morphologically Every argument of the verb is associated with a particular case assignment Ashwini Vaidya Participants 6 / 17 All the below are subjects, but are also agents, recipients, patients and possessors 1 He hit the ball 2 He received a blow 3 He received a gift 4 He has black hair There is a link between these semantically-typed verb arguments and their respective predicates Fillmore’s paper ‘The Case for Case’ termed these as case relations Case relations (in Fillmore’s sense) do not need to be realized morphologically Every argument of the verb is associated with a particular case assignment The case relations in the sentence are not inferable easily from syntactical relations alone Ashwini Vaidya Participants 6 / 17 Ram ne Mohan ko maara Notion of karta and karma The kind of case relations we are particularly interested in don’t need to be realized with a case-marker or case clitic Moreover, a one-to-one correspondence may be misleading e.g. Ram ko chaand dikhaa (ko will not always correspond to Patient Fillmore defined Agentive, Instrumental, Dative, Factitive (result), Locative and Objective case relations in the original paper Ashwini Vaidya Participants 7 / 17 Generalize over predicate types 1 John likes death metal (Dative-Objective) 2 Death metal pleases John (Objective-Dative) 1 John hits the ball 2 John breaks the glass 3 John cuts the cake All the above are Agentive and Objective If we are able to infer these relations between nouns, we can also infer the type of verb being used Ashwini Vaidya Participants 8 / 17 Generalize over different surface realizations 1 John broke the window [Agentive, Objective] Ashwini Vaidya Participants 9 / 17 Generalize over different surface realizations 1 John broke the window [Agentive, Objective] 2 The window broke [Objective] Ashwini Vaidya Participants 9 / 17 Generalize over different surface realizations 1 John broke the window [Agentive, Objective] 2 The window broke [Objective] 3 The window was broken by John [Objective, Agentive] Ashwini Vaidya Participants 9 / 17 Generalize over different surface realizations 1 John broke the window [Agentive, Objective] 2 The window broke [Objective] 3 The window was broken by John [Objective, Agentive] 4 The rock broke the window [Instrumental, Objective] Ashwini Vaidya Participants 9 / 17 Generalize over different surface realizations 1 John broke the window [Agentive, Objective] 2 The window broke [Objective] 3 The window was broken by John [Objective, Agentive] 4 The rock broke the window [Instrumental, Objective] The arguments of break can be realized in a myriad ways, but always as some combination of Agentive, Objective and Instrumental Ashwini Vaidya Participants 9 / 17 Generalize over different surface realizations 1 John broke the window [Agentive, Objective] 2 The window broke [Objective] 3 The window was broken by John [Objective, Agentive] 4 The rock broke the window [Instrumental, Objective] The arguments of break can be realized in a myriad ways, but always as some combination of Agentive, Objective and Instrumental We can make simple inferences from a shallow meaning representation e.g. in a document like Company A acquired Company B we can decide if it answers the question Was Company B acquired ? Ashwini Vaidya Participants 9 / 17 Role Description Examples Agent Initiator of action, capa- The batter smashed the pitch ble of volition into left field.. Patient Affected by action, un- John broke the window. dergoes change of state Theme Entity moving, or being Paola threw the Frisbee. ”located” Experiencer Perceives action but not He tasted the delicate flavor of in control the baby lettuce Beneficiary For whose benefit action He sliced me a large chunk of is performed prime rib. Instrument Intermediary He shot the wounded buffalo with a rifle Location Place of action The band played on the stage. Source Starting point The jet took off from Nairobi. Goal Ending point The ball rolled to the corner Ashwini Vaidya Participants 10 / 17 a. Der Hase frißt gleich den Kohl. (The hare-nom eats shortly the cabbage-acc) Ashwini Vaidya Participants 11 / 17 b. Den Hasen frißt gleich der Fuchs. (The hare-acc is eaten shortly the fox-nom) Ashwini Vaidya Participants 12 / 17 Kamide, Scheepers and Altmann (2003) German-speaking participants heard these two types of sentences while looking at the pictures shown earlier (visual world paradigm) Ashwini Vaidya Participants 13 / 17 Kamide, Scheepers and Altmann (2003) German-speaking participants heard these two types of sentences while looking at the pictures shown earlier (visual world paradigm) The authors wished to understand whether the German speakers would anticipate the target of the eating (cabbage in the first case) and (hare in the second case) as soon as they saw the first noun with case marker Der/Den Ashwini Vaidya Participants 13 / 17 Kamide, Scheepers and Altmann (2003) German-speaking participants heard these two types of sentences while looking at the pictures shown earlier (visual world paradigm) The authors wished to understand whether the German speakers would anticipate the target of the eating (cabbage in the first case) and (hare in the second case) as soon as they saw the first noun with case marker Der/Den Given the array of items in the picture, it would seem most likely that the participants would anticipate these as objects Ashwini Vaidya Participants 13 / 17 Kamide, Scheepers and Altmann (2003) German-speaking participants heard these two types of sentences while looking at the pictures shown earlier (visual world paradigm) The authors wished to understand whether the German speakers would anticipate the target of the eating (cabbage in the first case) and (hare in the second case) as soon as they saw the first noun with case marker Der/Den Given the array of items in the picture, it would seem most likely that the participants would anticipate these as objects Using eye-tracking, the authors showed that once the participants encountered the verb, looks towards the likely object increased Ashwini Vaidya Participants 13 / 17 Kamide, Scheepers and Altmann (2003) German-speaking participants heard these two types of sentences while looking at the pictures shown earlier (visual world paradigm) The authors wished to understand whether the German speakers would anticipate the target of the eating (cabbage in the first case) and (hare in the second case) as soon as they saw the first noun with case marker Der/Den Given the array of items in the picture, it would seem most likely that the participants would anticipate these as objects Using eye-tracking, the authors showed that once the participants encountered the verb, looks towards the likely object increased This shows that verbal semantic constraints along with case are integrated early during sentence processing Ashwini Vaidya Participants 13 / 17 Issues with thematic roles However, one challenge is that there’s no universally agreed upon set of roles Items with the “same” role (e.g., Instrument) may have more fine-grained meanings E.g. intermediary instruments can appear as subjects (1-2), but enabling instruments cannot (3-4) 1 The cook opened the jar with the new gadget. 2 The new gadget opened the jar. 3 Shelly ate the sliced banana with a fork. 4 *The fork ate the sliced banana. Are most Agents volitional ? E.g. A storm blew down a section of our fence or The tractor dragged the tree stump Hindi: maine ram-ko khaana khilaaya - What is the role for Ram ? Ashwini Vaidya Participants 14 / 17 Semantic frames As roles are too multifarious: one solution is to group them together Verbs denote aspects of scene:— scenes have participants. Verbs highlight some participants in a scene and background others 1 The price of bananas increased 5% 2 The price of bananas rose 5% 3 There has been a 5% rise in the price of bananas Ashwini Vaidya Participants 15 / 17 Semantic frames A frame defines frame-specific semantic roles, called frame elements + set of predicates/phrases that use these roles Commercial transaction frame BUYER buy GOODS (SELLER) (PRICE) subject object from for Angela bought the owl from Pete for 100 dollars Dan sold it to John He spent on a new computer 100 dollars Susan cost The telephone 100 dollars ∗ The last two sentences in their right order read: He spent 100 dollars on a new computer, The telephone cost Susan 100 dollars Ashwini Vaidya Participants 16 / 17 Frames will highlight or ‘profile’ certain participants depending upon the verb E.g in the ’Robbery’ frame, it’s possible to have two different verbs, rob and steal I Jesse robbed the rich (of all their money) Ashwini Vaidya Participants 17 / 17 Frames will highlight or ‘profile’ certain participants depending upon the verb E.g in the ’Robbery’ frame, it’s possible to have two different verbs, rob and steal I Jesse robbed the rich (of all their money) I ∗Jesse robbed a million dollars (from the rich) Ashwini Vaidya Participants 17 / 17 Frames will highlight or ‘profile’ certain participants depending upon the verb E.g in the ’Robbery’ frame, it’s possible to have two different verbs, rob and steal I Jesse robbed the rich (of all their money) I ∗Jesse robbed a million dollars (from the rich) I Jesse stole money Ashwini Vaidya Participants 17 / 17 Frames will highlight or ‘profile’ certain participants depending upon the verb E.g in the ’Robbery’ frame, it’s possible to have two different verbs, rob and steal I Jesse robbed the rich (of all their money) I ∗Jesse robbed a million dollars (from the rich) I Jesse stole money I ∗Jesse stole the rich (of money) Ashwini Vaidya Participants 17 / 17 Frames will highlight or ‘profile’ certain participants depending upon the verb E.g in the ’Robbery’ frame, it’s possible to have two different verbs, rob and steal I Jesse robbed the rich (of all their money) I ∗Jesse robbed a million dollars (from the rich) I Jesse stole money I ∗Jesse stole the rich (of money) In the cae of rob, victim and perpetrator are profiled/highlighted Ashwini Vaidya Participants 17 / 17 Frames will highlight or ‘profile’ certain participants depending upon the verb E.g in the ’Robbery’ frame, it’s possible to have two different verbs, rob and steal I Jesse robbed the rich (of all their money) I ∗Jesse robbed a million dollars (from the rich) I Jesse stole money I ∗Jesse stole the rich (of money) In the cae of rob, victim and perpetrator are profiled/highlighted In the case of steal the perpetrator and the valuables(source) are profiled Ashwini Vaidya Participants 17 / 17  Participants -II Ashwini Vaidya 21st Oct 2024 Ashwini Vaidya Participants -II 1 / 17 Semantic roles Helen drove to the party He swatted the fly with a newspaper The baboon was asleep on the roof of my car Joan drank the mug of ale Campbell saw the gun first George gave the doorman a tip List of themes: Agent, Patient, Theme, Experiencer,Beneficiary, Instrument, Goal, Source, Location Ashwini Vaidya Participants -II 2 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Experiencer: Entity aware of the action/state but not in control of it (Kevin felt ill) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Experiencer: Entity aware of the action/state but not in control of it (Kevin felt ill) Beneficiary: Entity for whose benefit the action was performed (They baked me a cake) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Experiencer: Entity aware of the action/state but not in control of it (Kevin felt ill) Beneficiary: Entity for whose benefit the action was performed (They baked me a cake) Instrument: Means by which the action comes about (She signed the document with a pen) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Experiencer: Entity aware of the action/state but not in control of it (Kevin felt ill) Beneficiary: Entity for whose benefit the action was performed (They baked me a cake) Instrument: Means by which the action comes about (She signed the document with a pen) Location: Where something is situated/takes place: (The band played in the pub) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Experiencer: Entity aware of the action/state but not in control of it (Kevin felt ill) Beneficiary: Entity for whose benefit the action was performed (They baked me a cake) Instrument: Means by which the action comes about (She signed the document with a pen) Location: Where something is situated/takes place: (The band played in the pub) Goal: Entity toward which something moves (Pat told the joke to her friends) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Agent: Initiator of the action (David cooked the omlette) Patient: Entity undergoing the effect of the action, undergoing a change of state (The sun melted the ice) Theme: Entity moved by the action/whose location is described: Roberto passed the ball Experiencer: Entity aware of the action/state but not in control of it (Kevin felt ill) Beneficiary: Entity for whose benefit the action was performed (They baked me a cake) Instrument: Means by which the action comes about (She signed the document with a pen) Location: Where something is situated/takes place: (The band played in the pub) Goal: Entity toward which something moves (Pat told the joke to her friends) Source: Entity from which something moves (The plane came back from Osaka) Ashwini Vaidya Participants -II 3 / 17 Semantic roles Helen drove to the party (Agent - Goal) Ashwini Vaidya Participants -II 4 / 17 Semantic roles Helen drove to the party (Agent - Goal) He swatted the fly with a newspaper (Agent, Patient, Instrument) Ashwini Vaidya Participants -II 4 / 17 Semantic roles Helen drove to the party (Agent - Goal) He swatted the fly with a newspaper (Agent, Patient, Instrument) The baboon was asleep on the roof of my car (Experiencer, Location) Ashwini Vaidya Participants -II 4 / 17 Semantic roles Helen drove to the party (Agent - Goal) He swatted the fly with a newspaper (Agent, Patient, Instrument) The baboon was asleep on the roof of my car (Experiencer, Location) Joan drank the mug of ale (Agent - Patient) Ashwini Vaidya Participants -II 4 / 17 Semantic roles Helen drove to the party (Agent - Goal) He swatted the fly with a newspaper (Agent, Patient, Instrument) The baboon was asleep on the roof of my car (Experiencer, Location) Joan drank the mug of ale (Agent - Patient) Campbell saw the gun first (Experiencer - Theme) Ashwini Vaidya Participants -II 4 / 17 Semantic roles Helen drove to the party (Agent - Goal) He swatted the fly with a newspaper (Agent, Patient, Instrument) The baboon was asleep on the roof of my car (Experiencer, Location) Joan drank the mug of ale (Agent - Patient) Campbell saw the gun first (Experiencer - Theme) George gave the doorman a tip (Agent, Beneficiary, Theme) Ashwini Vaidya Participants -II 4 / 17 In the following sentence: The police saw the robbers leaving the bank, the police are experiencers Can you think of sentences which have as their subjects the following:- I Agent I Patient I Theme I Recipient I Instrument Ashwini Vaidya Participants -II 5 / 17 Some possibilities:- John hit Peter Ashwini Vaidya Participants -II 6 / 17 Some possibilities:- John hit Peter The bowl cracked Ashwini Vaidya Participants -II 6 / 17 Some possibilities:- John hit Peter The bowl cracked The arrow flew through the air/ The glass fell Ashwini Vaidya Participants -II 6 / 17 Some possibilities:- John hit Peter The bowl cracked The arrow flew through the air/ The glass fell She received a demand for unpaid tax Ashwini Vaidya Participants -II 6 / 17 Some possibilities:- John hit Peter The bowl cracked The arrow flew through the air/ The glass fell She received a demand for unpaid tax The key opened the lock Ashwini Vaidya Participants -II 6 / 17 Thematic role grid Associated with each verb is some grammatical information: is it transitive/intransitive/ditransitive But also associated with each verb is knowledge of semantic roles We can associate each verb with a thematic role grid put V: < AGENT, THEME, LOCATION > This shows us that put is a three argument verb associated with certain thematic roles All of these roles are required– hence it’s not possible to say ∗John put the book We have to say John put the book on the table On the other hand, it’s possible to say John read the book:- because the theta grid for read has only two participant roles Ashwini Vaidya Participants -II 7 / 17 Participant roles and Non-participant roles Participant roles correspond to grammatical arguments: they are needed by the verb and must be obligatorily present Ashwini Vaidya Participants -II 8 / 17 Participant roles and Non-participant roles Participant roles correspond to grammatical arguments: they are needed by the verb and must be obligatorily present Non-participant roles are optional: they provide information about time, location, purpose or result of the event Ashwini Vaidya Participants -II 8 / 17 Participant roles and Non-participant roles Participant roles correspond to grammatical arguments: they are needed by the verb and must be obligatorily present Non-participant roles are optional: they provide information about time, location, purpose or result of the event For the purposes of the thematic role grids, only participant roles are relevant Ashwini Vaidya Participants -II 8 / 17 Participant roles and Non-participant roles Participant roles correspond to grammatical arguments: they are needed by the verb and must be obligatorily present Non-participant roles are optional: they provide information about time, location, purpose or result of the event For the purposes of the thematic role grids, only participant roles are relevant Verbs that share common thematic grids can form classes: E.g verbs of Transfer in English will contain examples like give, lend, supply, pay, donate, contribute Here, transfer is encoded from the perspective of the AGENT V: < AGENT, THEME, RECIPIENT > Barbara loaned money to Michael Ashwini Vaidya Participants -II 8 / 17 Another class of Transfer verbs will encode the transfer from the perspective of the RECIPIENT V: < RECIPIENT, THEME, SOURCE > Michael borrowed the money from Barbara Examples of this type include: Ashwini Vaidya Participants -II 9 / 17 Another class of Transfer verbs will encode the transfer from the perspective of the RECIPIENT V: < RECIPIENT, THEME, SOURCE > Michael borrowed the money from Barbara Examples of this type include: receive, accept, buy, purchase, rent, hire Ashwini Vaidya Participants -II 9 / 17 Voice Flexibility in semantic roles is offered by using the passive voice I Billy fed the horses (Active) I The horses were fed by Billy (Passive) CITY HALL DAMAGED BY STORMS Ashwini Vaidya Participants -II 10 / 17 Voice Flexibility in semantic roles is offered by using the passive voice I Billy fed the horses (Active) I The horses were fed by Billy (Passive) CITY HALL DAMAGED BY STORMS I In the sentence above, we have a past participle ‘damaged’, and the sentence is passive I However, the headline writer has chosen to omit the verb ‘be’ I Newspaper headlines allow passive forms to be written this way: however in most other contexts,passive form would be City hall was damaged by storms I To form the passive, we need both the participial form of the verb and the auxiliary (most commonly be) which provides tense information e.g. City hall is damaged by storms or City hall was damaged by storms Ashwini Vaidya Participants -II 10 / 17 In the active sentence, Billy is the AGENT and the subject, whereas horses is the PATIENT and object In the passive version, the PATIENT is the subject and the AGENT occurs as part of the prepositional phrase by Billy The passive sentence allows the speaker to describe the situation from the point of view of the PATIENT I Billy fed the horses (Active) I The horses were fed by Billy (Passive) A passive voice may be used to obscure the identity of the AGENT Ashwini Vaidya Participants -II 11 / 17 I was allowed to believe things that weren’t true Source: Language Log Ashwini Vaidya Participants -II 12 / 17 An agent may be so far backgrounded that it becomes an implied participant The horses were fed The foregrounding of the patient and demotion of the agent alters the perspective for the listener This choice of passive depends on what the speaker believes is more salient in the conversation Ashwini Vaidya Participants -II 13 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT Ashwini Vaidya Participants -II 14 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT The following examples allow THEME, RECIPIENT roles as subjects of passives:- Ashwini Vaidya Participants -II 14 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT The following examples allow THEME, RECIPIENT roles as subjects of passives:- I This money was donated to the school (Theme) Ashwini Vaidya Participants -II 14 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT The following examples allow THEME, RECIPIENT roles as subjects of passives:- I This money was donated to the school (Theme) I He was given a camera by his grandmother Ashwini Vaidya Participants -II 14 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT The following examples allow THEME, RECIPIENT roles as subjects of passives:- I This money was donated to the school (Theme) I He was given a camera by his grandmother Compare these to their active counterparts Ashwini Vaidya Participants -II 14 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT The following examples allow THEME, RECIPIENT roles as subjects of passives:- I This money was donated to the school (Theme) I He was given a camera by his grandmother Compare these to their active counterparts I Someone donated this money to the school Ashwini Vaidya Participants -II 14 / 17 Demotion of object Passives also allow the foregrounding of roles other than PATIENT The following examples allow THEME, RECIPIENT roles as subjects of passives:- I This money was donated to the school (Theme) I He was given a camera by his grandmother Compare these to their active counterparts I Someone donated this money to the school I His grandmother gave him a camera Ashwini Vaidya Participants -II 14 / 17 However, the general pattern is that the grammatical object is always made the subject The recipient in the example below (her) is an indirect object, whereas garage is a direct object John built a garage for her ∗For her was built a garage by John A garage was built for her by John We can also see this in the example He opened the door with this key ∗With this key was opened the door by him Ashwini Vaidya Participants -II 15 / 17  In the following sentences is the rule followed? Passivize the following: 1 He loaded hay on to the tractor 2 He loaded the tractor with hay Ashwini Vaidya Participants -II 16 / 17  In the following sentences is the rule followed? Passivize the following: 1 He loaded hay on to the tractor 2 He loaded the