Lexical Semantics CMSC 723 / LING 723 / INST 725 MARINE CARPUAT.

Transcription

1 Lexical Semantics CMSC 723 / LING 723 / INST 725 MARINE CARPUAT marine@cs.umd.edu

2 Q: What s meaning? Compositional semantics answer:

3 Q: What s meaning? We now answer from a lexical semantics perspective i.e. let s think about the meaning of words

4 Today Word senses & word relations WordNet Word Sense Disambiguation Supervised Semi-supervised

5 WORD SENSES & WORD RELATIONS

6 Where can we look up the meaning of words? Dictionary?

7 Word Senses Word sense = distinct meaning of a word Same word, different senses Homonyms (homonymy): unrelated senses; identical orthographic form is coincidental Polysemes (polysemy): related, but distinct senses Metonyms (metonymy): stand in, technically, a subcase of polysemy Different word, same sense Synonyms (synonymy)

8 Just to confuse you Homophones: same pronunciation, different orthography, different meaning Examples: would/wood, to/too/two Homographs: distinct senses, same orthographic form, different pronunciation Examples: bass (fish) vs. bass (instrument)

9 Relationship Between Senses IS-A relationships From specific to general (up): hypernym (hypernymy) From general to specific (down): hyponym (hyponymy) Part-Whole relationships wheel is a meronym of car (meronymy) car is a holonym of wheel (holonymy)

10 WordNet: a lexical database for English Includes most English nouns, verbs, adjectives, adverbs Electronic format makes it amenable to automatic manipulation: used in many NLP applications WordNets generically refers to similar resources in other languages

11 WordNet: History Research in artificial intelligence: How do humans store and access knowledge about concept? Hypothesis: concepts are interconnected via meaningful relations Useful for reasoning The WordNet project started in 1986 Can most (all?) of the words in a language be represented as a semantic network where words are interlinked by meaning? If so, the result would be a large semantic network

12 Synonymy in WordNet WordNet is organized in terms of synsets Unordered set of (roughly) synonymous words (or multi-word phrases) Each synset expresses a distinct meaning/concept

13 WordNet: Example Noun {pipe, tobacco pipe} (a tube with a small bowl at one end; used for smoking tobacco) {pipe, pipage, piping} (a long tube made of metal or plastic that is used to carry water or oil or gas etc.) {pipe, tube} (a hollow cylindrical shape) {pipe} (a tubular wind instrument) {organ pipe, pipe, pipework} (the flues and stops on a pipe organ) Verb {shriek, shrill, pipe up, pipe} (utter a shrill cry) {pipe} (transport by pipeline) pipe oil, water, and gas into the desert {pipe} (play on a pipe) pipe a tune {pipe} (trim with piping) pipe the skirt What do you think of the sense granularity?

14 The Net Part of WordNet {conveyance; transport} hyperonym {vehicle} hyperonym {bumper} {motor vehicle; automotivevehicle} meronym hyperonym {car dor} {car window} {car; auto; automobile; machine; motorcar} {armrest} meronym hyperonym hyperonym {cruiser; squadcar; patrol car; policecar; prowl car} {cab; taxi; hack; taxicab; } {car miror} meronym meronym {hinge; flexiblejoint} {dorlock}

15 WordNet: Size Part of speech Word form Synsets Noun 117,798 82,115 Verb 11,529 13,767 Adjective 21,479 18,156 Adverb 4,481 3,621 Total 155, ,659

16 Word Sense From WordNet: Noun {pipe, tobacco pipe} (a tube with a small bowl at one end; used for smoking tobacco) {pipe, pipage, piping} (a long tube made of metal or plastic that is used to carry water or oil or gas etc.) {pipe, tube} (a hollow cylindrical shape) {pipe} (a tubular wind instrument) {organ pipe, pipe, pipework} (the flues and stops on a pipe organ) Verb {shriek, shrill, pipe up, pipe} (utter a shrill cry) {pipe} (transport by pipeline) pipe oil, water, and gas into the desert {pipe} (play on a pipe) pipe a tune {pipe} (trim with piping) pipe the skirt

17 WORD SENSE DISAMBIGUATION

18 Word Sense Disambiguation Task: automatically select the correct sense of a word Lexical sample All-words Motivated by many applications: Semantic similarity Information retrieval Machine translation But gains in performance have been elusive.

19 How big is the problem? Most words in English have only one sense 62% in Longman s Dictionary of Contemporary English 79% in WordNet But the others tend to have several senses Average of 3.83 in LDOCE Average of 2.96 in WordNet Ambiguous words are more frequently used In the British National Corpus, 84% of instances have more than one sense Some senses are more frequent than others

20 Ground Truth Which sense inventory do we use? Issues there? Application specificity?

21 Existing Corpora Lexical sample line-hard-serve corpus (4k sense-tagged examples) interest corpus (2,369 sense-tagged examples) All-words SemCor (234k words, subset of Brown Corpus) Senseval/SemEval (2081 tagged content words from 5k total words)

22 Evaluation Intrinsic Measure accuracy of sense selection wrt ground truth Extrinsic Integrate WSD as part of a bigger end-to-end system, e.g., machine translation or information retrieval Compare WSD

23 Baseline Performance Baseline: most frequent sense Equivalent to take first sense in WordNet Does surprisingly well! 62% accuracy in this case!

24 Upper Bound Performance Upper bound Fine-grained WordNet sense: 75-80% human agreement Coarser-grained inventories: 90% human agreement possible

25 WSD Approaches Depending on use of manually created knowledge sources Knowledge-lean Knowledge-rich Depending on use of labeled data Supervised Semi- or minimally supervised Unsupervised

26 Simplest WSD algorithm: Lesk s Algorithm Intuition: note word overlap between context and dictionary entries Unsupervised, but knowledge rich The bank can guarantee deposits will eventually cover future tuition costs because it invests in adjustable-rate mortgage securities. WordNet

27 Lesk s Algorithm Simplest implementation: Count overlapping content words between glosses and context Lots of variants: Include the examples in dictionary definitions Include hypernyms and hyponyms Give more weight to larger overlaps (e.g., bigrams) Give extra weight to infrequent words (e.g., idf weighting)

28 Another approach: Supervised WSD How would you frame WSD as a supervised classification task?

29 Supervised Classification Training Testing training data? unlabeled document label 1 label 2 label 3 label 4 supervised machine learning algorithm Representation Function Classifier label 1? label 2? label 3? label 4?

30 Features Possible features POS and surface form of the word itself Surrounding words and POS tag Positional information of surrounding words and POS tags Same as above, but with n-grams Grammatical information Richness of the features? Richer features = ML algorithm does less of the work More impoverished features = ML algorithm does more of the work

31 Classifiers Once we cast the WSD problem as supervised classification, any multiclass classifer can be used: Perceptron Support Vector Machines

32 WSD Accuracy Generally Supervised approaches yield ~70-80% accuracy However depends on actual word, sense inventory, amount of training data, number of features, etc.

33 WORD SENSE DISAMBIGUATION: MINIMIZING SUPERVISION

34 Minimally Supervised WSD Problem: annotations are expensive! Solution 1: Bootstrapping or co-training (Yarowsky 1995) Start with (small) seed, learn classifier Use classifier to label rest of corpus Retain confident labels, treat as annotated data to learn new classifier Repeat Heuristics (derived from observation): One sense per discourse One sense per collocation

35 One Sense per Discourse A word tends to preserve its meaning across all its occurrences in a given discourse Evaluation: 8 words with two-way ambiguity, e.g. plant, crane, etc. 98% of the two-word occurrences in the same discourse carry the same meaning the grain of salt: accuracy depends on granularity Performance of one sense per discourse measured on SemCor is approximately 70%

36 One Sense per Collocation A word tends to preserve its meaning when used in the same collocation Strong for adjacent collocations Weaker as the distance between words increases Evaluation: 97% precision on words with two-way ambiguity Again, accuracy depends on granularity: 70% precision on SemCor words

37 Yarowsky s Method: Example Disambiguating plant: industrial sense vs. living thing sense Think of seed features for each sense Industrial sense: co-occurring with manufacturing Living thing sense: co-occurring with life Use one sense per collocation to build initial decision list classifier Treat results as annotated data, train new decision list classifier, iterate

38 Decision List Ordered list of tests (equivalent to case statement): Example decision list, discriminating between bass (fish) and bass (music) :

39 Building Decision Lists Simple algorithm: Compute how discriminative each feature is: log P( S P( S 1 2 ) ) Create ordered list of tests from these values f f i i

40

41

42

43

44 Yarowsky s Method: Stopping Stop when: Error on training data is less than a threshold No more training data is covered Use final decision list for WSD

45 Yarowsky s Method: Discussion Advantages: Accuracy is about as good as a supervised algorithm Bootstrapping: far less manual effort Disadvantages: Seeds may be tricky to construct Works only for coarse-grained sense distinctions Snowballing error with co-training

46 WSD with Parallel Text Problems: annotations are expensive! What s the proper sense inventory? How fine or coarse grained? Application specific? Observation: multiple senses translate to different words in other languages! Use the foreign language as the sense inventory! Added bonus: annotations for free! (Byproduct of word-alignment process in machine translation)

47 Today Word senses & word relations WordNet Word Sense Disambiguation Lesk Algorithm Supervised classification Semi-supervised approaches Co-training: combine 2 views of data Use parallel text Next: semantic similarity & more unsupervised learning