Intelligent Access to Text: Integrating Information Extraction Technology into Text Browsers

Intelligent Access to Text: Integrating Information Extraction Technology into Text Browsers Robert Gaizauskas � Patrick Herring � Michael Oakes � Michelline Beaulieu � Peter Willett � Helene Fowkes � Anna Jonsson � � Department of Computer Science / � Department of Information Studies University of Sheffield Regent Court, Portobello Road Sheffield S1 4DP UK � initial.surname � @sheffield.ac.uk ABSTRACT In this paper we show how two standard outputs from information extraction (IE) systems – named entity annotations and scenario templates – can be used to enhance access to text collections via a standard text browser. We describe how this information is used in a prototype system designed to support information workers’ ac- cess to a pharmaceutical news archive as part of their “industry watch” function. We also report results of a preliminary, qualita- tive user evaluation of the system, which while broadly positive in- dicates further work needs to be done on the interface to make users aware of the increased potential of IE-enhanced text browsers. 1. INTRODUCTION Information extraction (IE) technology, as promoted and defined by the DARPA Message Understanding Conferences [4, 5] and the current ACE component of TIDES [1], has resulted in impres- sive new abilities to extract structured information from texts, and complements more traditional information retrieval (IR) technol- ogy which retrieves documents or passages of relevance from text collections and leaves information seekers to browse the retrieved sub-collection (e.g. [2]). However, while IR technology has been readily incorporated into end-user applications (e.g. web search engines), IE technology has not yet been as successfully deployed in end-user systems as its proponents had hoped. There are several reasons for this, including: 1. Porting cost. Moving IE systems to new domains requires considerable expenditure of time and expertise, either to cre- ate/modify domain-specific resources and rule bases, or to annotate texts for supervised machine learning approaches. 2. Sensitivity to inaccuracies in extracted data. IE holds out the promise of being able to construct structured databases from text sources automatically, but extraction results are by no means perfect. Thus, the technology is only appropriate . for applications where some error is tolerable and readily de- tectable by end users. 3. Complexity of integration into end-user systems. IE sys- tems produce results (named entity tagged texts, filled tem- plates) which must be incorporated into larger, more sophis- ticated application systems if end users are to gain benefit from them. In this paper we present the approach taken in the TRESTLE project (Text Retrieval Extraction and Summarisation Technolo- gies for Large Enterprises) which addresses the second and third of these problems; and also preliminary results from the user testing evaluation of the TRESTLE interface. The goal of the TRESTLE project is to develop an advanced text access facility to support information workers at GlaxoSmithKline (GSK), a large pharma- ceutical corporation. Specifically, the project aims to provide en- hanced access to Scrip1, the largest circulation pharmaceutical in- dustry newsletter, in order to increase the effectiveness of employ- ees in their “industry watch” …

Content truncated to 500 words