|
Probabilistic Models of Language Structure
Chris Manning
Department of Linguistics
Sydney University
When: Friday 2nd May, 1997
Time: 2:00pm
Where: Room E6A102, Macquarie University
Abstract:
In Natural Language Processing, the 1990s have seen an enormous swing away from (hand-crafted) knowledge-based systems towards the use of Statistical and Corpus-based Natural Language Processing methods. In this talk, I will motivate this approach, showing how it has been driven by a wish to deal with `real world' problems, but also how it actually has considerable linguistic interest. I will then review the strengths and weaknesses of the most used probabilistic models of langauge (that is, word n-gram models, and probabilistic context free grammars). Thereafter, I will look at some models for probabilistic parsing that I have been exploring recently. These cover models with more structural context, binarized grammars that seek to remedy some of the biases of probabilistic context free grammars, and dependency/subcategorization-based approaches, which (re)introduce lexical content.
Enquiries: sals@mri.mq.edu.au
| Last modified: July 1997 |