Methods for quality assurance in semi-automatic lexicon acquisition from corpora
Judith Eckle‐Kohler · 1998
This paper presents linguistics-based methods and engineering methods for quality assurance in semi-automatic acquisition of broad coverage lexicons from corpora. Automated linguistic tests are used to acquire candidates for particular subcategorization frames automatically; the regular use of metrics in the acquisition process contributes to a controlled development of these tests. The proposed methods are illustrated by the acquisition of a particular class of verbs taking daß-clauses in German, showing how the precision of the automatically acquired data can be maximized with only a slight decrease in recall. Keywords: Quality assurance in lexicon acquisition, semi-automatic, corpus-based lexicon acquisition. 1. Introduction 1.1 Motivation Quality assurance (QA in the following) has been recognized as an integral part of lexicon acquisition, not only in large-scale manual acquisition as in COMLEX (cf. Grishman et al. (1994)), but also in automatic acquisition from corpora as descr...