UC BERKELEY
EECS technical reports
TECHNICAL REPORTS


CSD-01-1148.pdf
CSD-01-1148.ps
Conditions of Use

Archive Home Page

Cubic-time Parsing and Learning Algorithms for Grammatical Bigram Models

Authors:
Paskin, Mark A.
Technical Report Identifier: CSD-01-1148
June 20, 2001
CSD-01-1148.pdf
CSD-01-1148.ps

Abstract: This paper presents a probabilistic model of English grammar that is based upon "grammatical bigrams", i.e., syntactic relationships between pairs of words. Because of its simplicity, the grammatical bigram model admits cubic-time parsing and unsupervised learning algorithms, which are described in detail.