Adaptive Information Extraction and Sublanguage Analysis
Ralph Grishman · 2001
Introduction 1 Information extraction (IE) has made significant progress in the last decade. We have developed practical, efficient approaches to IE which have yielded modest levels of performance on general texts and quite good performance on restricted, `semi-structured' texts. More notably, over the last few years there has been a blossoming of work in adaptive IE --- the topic of this and other recent workshops --- IE systems which can be rapidly and automatically (or semi-automatically) moved to new extraction tasks. To date, these developments have been relatively little influenced by linguistic studies of the texts. In fact, the trend has been towards less linguistic analysis. Some early IE systems used full parsing and in a few cases relatively deep semantic analysis. Because of limitations of full parsing methods (particularly a decade ago) this gave way to a common methodology based on limited parsing and simple pattern matching. Adaptive IE systems have in