Tailoring a Broad-Coverage System for the Analysis of Dictionary Definitions

Abstract

Machine readable dictionaries are being exploited by automatically extracting semantic knowledge from their definitions in order to construct lexical data and/or knowledge bases. This chapter focuses on the extraction process of computing a syntactic analysis for each dictionary definition. Although the output of a broad-coverage grammar is adequate to make this analysis, we decided to add a small post-processing component, whose tasks are to rule out ambiguities and to reshape the initial analysis where necessary, according to peculiarities of the dictionary text structure. This separation allows us to revise the post-processor without disturbing the grammar, and to catalogue the actual differences between dictionary and general text.