Natural Language Models and Interfaces (Part B)


Lecturers: Ivan Titov (Part B) and Khalil Sima'an (Part A)
Instructors: Kristin Rieping (k.rieping at gmail.com) and Robrecht Jurriaans (robrechtjurriaans at gmail.com)

Short Description

Part B of the class taught by Ivan Titov can be regarded as a minicourse on syntactic parsing of natural language. We will look into syntactic representations for natural language, models of syntax (CFGs/PCFGs induced from annotated data), algorithms for predicting syntactic representations (CKY), problems with such treebank grammars and ways how to enhance treebank grammars (vertical and horizontal Markovization, lexicalization). In the practical part of the class you will be provided with a baseline (not fully functional) parser which you will modify to implement some of the ideas we discuss in class.


Reading

We will continue using the Jurafsky and Martin book (now chapters 12 -- 14). Sections / chapters related to specific lectures are listed below.


Grading (Part B)


Practical assignments

The practical assignments should be delivered on Blackboard for each of the two stages:

  • Stage 1: Computing the probability of a tree -- deadline May 18, fallback May 20 (code and 1-page report)
  • Stage 2: Adding support of unary rules in CKY (2.1) and grammar annotation (2.2) -- deadline May 26 May 31, fallback May 29 June 2 (code and 4 page final report). Implementation of grammar annotation part (stage 2.2) is not required anymore, but you still need to discuss grammar annotation techniques in your report (at least vertical and horizontal Markovization).
  • See the description for details. Please follow the same procedure as for Part A to submit your reports and results. The data and the baseline code are downloadable from Blackboard, make sure you can access them.
    Lectures

    May 6 Guest lecture by Raquel Fernandez
    May 13 Syntax: Introduction and basics Reading: J&M 12.1 - 12.4, 14.1, optionally 12.5 - 7
    May 17 Syntax: Parsing with (P)CFGs (CKY), parser evaluation Reading: J&M 13.1 - 13.4, 14.2 - 14.4, 14.7
    May 21 Syntax: Problems of PCFGs, grammar annotations Reading: J&M 14.1 - 14.6
    May 24 Final lecture for Parts A & B (Khalil Sima'an)

    Lectures slides will be made downloadable (after each lecture).