|
Title:
|
Towards 'PATR-DOP': Implementing a Stochastic Parser for Phrase Disambiguation
|
|
Author:
|
Neil Thompson
|
|
Email:
|
click here to access email
|
|
Degree Awarded:
|
University of Essex
, MA Computational Linguistics
|
|
Degree Date:
|
2002
|
|
Linguistic Subfield(s):
|
Computational Linguistics
|
|
Director(s):
|
Doug Arnold
|
|
|
Abstract:
|
|
|
Data oriented parsing is a technique which is able to assign probabilities to data in a corpus at many different levels. In this dissertation an implementation in Prolog obtains probabilities of tree structures and feature structures and uses these to find the most probable representation for a new input sentence. A simple top down parser is found to perform relatively well in this environment. Two techniques - Monte Carlo parsing and a proposed complementary derivation-restricting technique are used to reduce the data load and parse times.
|
|
|
E-mail this message to a friend
