LOT Winter School 2009

Course title

Computational Linguistics

Teacher

Gosse Bouma

 

Course Level:

Intermediate

The course is intended for linguists with an interest in computational applications and corpus linguistics.
It is assumed that the participants have a background in linguistics, but no specific knowledge of corpus or computational linguistics is presupposed.

 

Course Description

In recent years, ever larger text corpora have been created, which are of interest both for linguistic research, but also as data sources for practical applications. In this course, we give an overview of methods for automatically enriching such corpora with linguistic information, in particular with syntactic structure. We will discuss examples of linguistic research which is based on information obtained from such corpora. We will also show how syntactic structure can be used to extract various kinds of information from large text collections, and how such information can be used in applications.

 

Day-to-day Program

Monday: Creating Annotating, and Searching Corpora
 

This is a brief introduction in corpus linguistics. We discuss issues as size, representativeness, tokenization, Part of Speech tagging, Syntactic information, methods for manual and automatic annotation, formats, and search tools. We argue that properly annotated corpora are of interest for linguistic research as well as applications in the field of information search and extraction.

• Lecture Notes (4 on 1 A4)
• Stanford Parser
• BNC
• Gosse Bouma and Begona Villada (2001), Corpus-based acquisition of collocational prepositional phrases
• Grefenstette and Nioche (2000), Estimation of English and non-English language use on the WWW
• Oostendorp and van der Wouden, Corpus Internet
• Veronis, Google's counts faked?
• Liberman, Google-sampling: avoiding pseudo-text in cyberspace

 

Tuesday: Statistical Parsing

Automatic syntactic annotation requires accurate and robust methods for automatic analysis of text. We discuss various approaches to creating large grammars, ranging from systems that are the mainly created to systems that are induced automatically from corpus fragments.

• Lecture Notes
• Gertjan van Noord's research

 

Wednesday: Syntactic Patterns in Large Corpora
 

Syntactically annotated corpora make it possible to investigate the frequency of syntactic constructions, and of lexical items in specific syntactic constructions. We give an overview of corpus linguistic and psycholinguistic research in which such information has been used.

• Lecture Notes (4 on 1 A4)
• Bouma and Spenader (2009), The Distribution of Weak and Strong Object Reflexives in Dutch
• Bouma, Hendriks, and Hoeksema, Focus Particles inside Prepositional Phrases: A Comparison between Dutch, English, and German
• Bastiaanse, Bouma, Post, Linguistic Complexity and Frequency in Agrammatic Speech Production (draft)
• Dissertation by Gerlof Bouma on Word Order in Dutch

 

Thursday: Mining Syntactically Annotated Corpora

Various text mining applications may benefit from access to syntactic information. We will discuss the role of syntactic annotation in lexical acquisition, information extraction, and question answering.

• Lecture Notes (4 on 1 A4)
• Simar Words Demo
• Dissertation by Lonneke van der Plas on similar words

 

Friday: Coreference Resolution
 

Many of the research questions in corpus linguistics require semantic distinctions (i.e. thematic roles, information about the animacy of arguments, word sense) that are difficult to obtain using syntactic information only. Many practical applications can benefit from semantic information as well (i.e. named-entity classes, co-reference resolution, word sense, discourse relations).

• Lecture Notes (4 on 1 A4)
• Dissertation by Jori Mur on Coreference and Question Answering