How Advances in Search Combine Databases, Sentence Diagramming, and "Just the Facts"

Abstract

Among the most studied subfields of natural language processing (NLP) are information retrieval (IR, aka search) and information extraction (IE). Yet, even though researchers have wanted to combine IR and IE for nearly a decade, we are only now starting to see commercial applications that put the notion into practice. The search market remains dominated by traditional IR technologies, as evidenced by the likes of Google, Yahoo, and Verity. (Granted, Google's page-rank algorithm departs from traditional IR algorithms by including link data, but the underlying index of terms is conventional.) Two forces are driving the need for advanced search technologies: (a) users are more knowledgeable and are demanding greater search engine precision; and (b) vendors must create new search products, particularly enterprise search applications. This issue's theme articles represent several of these advancements. In this article, the author detail how IE-in particular action-oriented event extraction-enables new search functionality