arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Artificial Intelligence

Automatic extraction of subcategorization frames for

Paula Chesley

See on the map

Abstract

This paper describes the integration of corpus-based syntactic subcategorization frames into a large-scale, theory-neutral lexical resource for French (Romary et al. (2004)). This database is the first to implement the Lexical Markup Framework (LMF), an international initiative towards ISO standards for lexical databases (ISO TC 37/SC 4). The subcategorization frames have been acquired via a dependency-based parser (Bick (2003)), whose verb lexicon is currently incomplete with respect to subcategorization frames. Therefore, we have implemented probabilistic filtering as a post-parsing step. Similar methods have proven efficient for English (Brent (1992), Manning (1993), Briscoe and Carroll (1997)) and other languages (Sarkar and Zeman (2000)). On a test set of 104 frequent verbs, most of which are in the Test Suites for Natural Language Processing for French (Lehmann et al. (1996)), we achieve lower bounds on type precision at 86.8 % and on token recall at 54.3%. Additionally, our work demonstrates that using a language resource, such as the VISL parser, with a currently unevaluated (and potentially high) error rate can yield robust results in conjunction with probabilistic filtering of the resource output. For languages currently lacking robust parsers and large-scale treebanks, such findings are favorable indeed. We created a multi-genre corpus of 200 random occurrences of the 104 verbs using the Frantext online literary database 1. We then had this corpus parsed with the VISL parser. As opposed to many works on subcategorization frame extraction, we do not assume a set of a priori frames to be extracted before filtering. Rather, the frames are determined throughout the filtering stage. However, we limit the syntactic categories treated to the following: • direct objects; • prepositional phrases (PPs) headed by a particular preposition; • subordinate clauses and small clauses with various markers; • raising and control verbs; • predicative adjectival phrases; • reflexive verbs. The resulting subcategorization frames can consist of any combination of the above categories.

More like this