1 / 31

LING 581: Advanced Computational Linguistics

LING 581: Advanced Computational Linguistics. Lecture Notes February 16th. Administrivia. Homework presentations today A related experiment … New topic: computational semantics. Homework. WSJ corpus : sections 00 through 24 Training : normally 02-21 (20 sections )

hasana
Download Presentation

LING 581: Advanced Computational Linguistics

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. LING 581: Advanced Computational Linguistics Lecture Notes February 16th

  2. Administrivia • Homework presentations today • A related experiment … • New topic: computational semantics

  3. Homework • WSJ corpus: sections 00 through 24 • Training: normally 02-21 (20 sections) • concatenate mrg files as input to Bikel Collins train • training is fast • Evaluation: on section 23 • use tsurgeon to get data ready for EVALB • use Bikel Collins parse repeatedly on wsj-23.lsp • 2400+ sentences, parsing is slow • Question: • How does the Bikel Collins vary in precision and recall? if you randomly pick 1, 2, 3 up to 20 sections to do the training with… • Use EVALB • plot precision and recall graphs

  4. Perl Code • Generate training data

  5. Training Data

  6. Training Data

  7. An experiment

  8. Parsing Data • All 5! (=120) permutations of • colorless green ideas sleep furiously .

  9. Parsing Data • The winning sentence was: • Furiously ideas sleep colorless green . • after training on sections 02-21 (approx. 40,000 sentences) sleep takes ADJP object with two heads adverb (RB) furiously modifies noun

  10. Parsing Data • The next two highest scoring permutations were: • Furiously green ideas sleep colorless . • Green ideas sleep furiously colorless . sleep takes NP object sleep takes ADJP object

  11. Parsing Data • (Pereira 2002) compared Chomsky’s original minimal pair: • colorless green ideas sleep furiously • furiously sleep ideas green colorless Ranked #23 and #36 respectively out of 120

  12. Parsing Data But … • graph (next slide) shows how arbitrary these rankings are: • furiously sleep ideas green colorless Vastly outranks • colorless green ideas sleep furiously (and the top 3 sentences) when trained on randomly chosen sections covering 14,188 to 31,616 WSJ PTB sentences

  13. Rank of Best Parse for Sentence Permutation Sentence

  14. Rank of Best Parse for Sentence Permutation • Note also that: • colorless green ideas sleep furiously totally outranks the top 3 (and #36) just briefly on one data point (when trained on 33605 sentences)

  15. New Topic Computational Semantics • (Portner 2005), an intro book titled What is meaning?, asks on page 1 • Why formalize? • can construct precise theories • precise theories are better • “they don’t allow the theorists to fudge the data quite so easily as less precise theories do” • Why implement?

  16. Formalization • Challenge: • we all understand the argument made on the previous page? • How do we formalize it? • Why formalize? • can construct precise theories • precise theories are better • “they don’t allow the theorists to fudge the data quite so easily as less precise theories do”

  17. Implementation • We can implement in Prolog, see next time … (Make sure you have SWI Prolog up and running for next time) • free download from www.swi-prolog.org

  18. Simple past vs. present perfect • (1) Mary received the most votes in the election • (2) Mary has received the most votes in the election • (so) Mary will be the next president • Idea • (1) reports a past event • (2) reports a past event and a current “result” • present perfect • expectation... • (entailment)

  19. Simple past vs. present perfect • (3) Will Mary be able to finish Dos Passos’USA trilogy by the next club meeting? It’s so long! • Well, she has read Remembrance of Things Past, and its even longer • what’s a “result” here? • expectation... • (entailment) • Mary has read a really long book before and therefore...

  20. Meaning • What is a Meaning? • difficult sometimes to pin down precisely • by reference to other words • foreign language: 犬(inu) = “dog”

  21. Meaning • What is a Meaning? • Example: • important • Merriam-Webster (sense 1): • marked by or indicative of significant worth or consequence:valuable in content or relationship

  22. Question • What is a Meaning? • Example: • important • Thesaurus • Text: 1 having great meaning or lasting effect • <the discovery of penicillin was a very important event in the history of medicine> • Synonyms big, consequential, eventful, major, material, meaningful, momentous, significant, substantial, weighty • Related Words decisive, fatal, fateful, strategic; earnest, grave, serious, sincere; distinctive, exceptional, impressive, outstanding, prominent, remarkable; valuable, worthwhile, worthy; distinguished, eminent, great, illustrious, preeminent, prestigious; famous, notorious, renowned; all-important, critical, crucial

  23. Question • What is a Meaning? • Meaning = Concept (or thought or idea) • “dog” maps to DOG • <word> maps to <concept> • Problems • need to provide a concept for every meaningful piece of language • how about expressions “whatever”, “three” • need to map different expressions into same concept • Externalization? Putnam’s twin earth experiment (H2O vs. XYZ) • DOG a shared concept?

  24. Question • What is a Meaning? • Meaning = Concept (or thought or idea) • twin earth experiment • same except H2O = XYZ • “water” refers to H2O • “water” refers to XYZ • identical twins on the two earths don’t mean the same thing by the word “water”

  25. Question • What is a Meaning? • Meaning = Concept (or thought or idea) • Skip the “Meaning = Concept” definition • reason the word “dog” means the same thing for you and me • not that we have the same mental constructs relating to the word • it’s because of our intention to apply the word “dog” to the same things out there in our environment

  26. Truth Conditions • The circle is inside the square • Can draw a picture of scenarios for which the statement is true and the statement is false • truth-conditions different from truth-value

  27. Truth Conditions • The circle is inside the square • Proposition expressed by a sentence is its truth-conditions • i.e. sets of possible worlds (aka situations)

  28. Truth Conditions • The circle is inside the square and the circle is dark • and = set intersection • Mary is a student and a baseball fan

  29. Truth Conditions • Mary and John bought a book • and = set intersection ? are Mary and John sets? how about “and = set union”?

  30. Truth Conditions • The square is bigger than the circle • The circle is smaller than the square • Given these two sentences, evaluate • Synonymous • (what does this mean wrt truth conditions?) • Contrary • Entailment • Tautology

  31. Practice • 1. Does sleep entail snore? • 2. Does snore presuppose sleep? • 3. Given the statement “All crows are black”, give an example of a sentence expressing a tautology involving this statement?

More Related