HG7021: Computational Grammars

Instructor Francis Bond. 2020, 2014, 2012.

HSSK MR4 Thursday 13:00-17:00

In this course we implement a computational grammar of a language (one each). You must have a laptop, preferably running Linux, or install a virtual machine running linux (I am most familiar with Ubuntu).

Course Outline

WeekTopicLabWikiReading
1 Overview, Introduction
Installing the LKB and dependencies Sag, Wasow and Bender (2003) Ch 1-2
2 The Grammar Matrix: Motivations, technical details Getting to know the LKB and the Grammar Matrix; Choose language Copestake (2002) Ch 1-3
3 Testsuites, [incr tsdb()] Testsuites/customization I: Word order, person/number/gender, pronouns, case, the rest of the NP, basic lexicon (full form) Copestake (2002) Ch 4, 5
(Oepen & Flickinger 1998)
4 Typed Feature Structures and Unification Testsuites/customization II: Tense/aspect, agreement, other required affixes, negation, argument optionality, demonstratives, Bender, Flickinger & Oepen (2002)
5 Minimal Recursion Semantics Testsuites/customization III: Matrix yes-no questions, coordination, modification, non-verbal predicates, embedded clauses, information structure Copestake, Flickinger, Pollard and Sag (2005)
6 Valence and Semantic Composition in HPSG; Chart Parsing Revision Git for Grammar Engineers Sag, Wasow and Bender (2003) ch 4-5
7 Modification, Discourse Status, Argument Optionality; Precision grammars and corpus data Modification, demonstratives, argument optionality Borthen and Haugereid (2005)
(Baldwin et al (2005))
9 Composition and Decomposition: Demonstrative pronouns and multi-word expressions Breaking words down and building phrases up Sag et al. (2002)
9 Clause types, illocutionary force Polar questions, embedded clauses, non-verbal predicates Sag, Wasow and Bender (2003) Ch 5
10 Negation, Raising, Control, Argument Composition I can eat glass. It doesn't hurt me. Sag, Wasow and Bender (2003) Ch 12
11 Treebanking Treebanking and Parse ranking FFTB Oepen, Flickinger and Bond (2004)
12 Deep Semantic Transfer: The LOGON MT architecture Information Structure and VPM Grammar clean up; Transfer rules VPM Bond et al. 2011
13 No Slides Classifiers

Textbooks and Tools

Tools

The Grammar Engineering Cycle

  1. Create/improve/degrade! a grammar
  2. Regression test: Parse the functional test suite and a general test suite
  3. If there was an improvement, commit it
    else back to step 1
  4. Treebank the new profiles (maybe after several improvements)
    The update function should make this cheap
  5. Identify and analyse your next phenomenon
  6. back to step 1

Readings

Assessment

Due 2020-04-28

Submission Instructions

From week 5 (when we have stopped making changes via the Matrix), please upload your grammar to github.

I recommend you add to .gitattributes

    *.odt binary
    *.pdf binary
  
This will mean git does not try to save diffs.
Concrete steps (first time only)
  • Concrete steps (subsequently)

    Acknowledgment

    Course materials borrow heavily from Linguistics 567: Knowledge Engineering for NLP at the University of Washington. Thanks to Emily Bender for letting us use them.

    Chart parsing slides by Ewan Kline, from the University of Edinburgh's Introduction to Computational Linguistics (2006/07)

    These slides are hosted on github at https://github.com/bond-lab/Grammar-Engineering.