WEBCOOP: A Cooperative Question Answering System on the Web

WEB COOP: A Cooperative Question-Answering System on the WEB Farah Benamara� Patrick Saint Dizier IRIT� IRIT Toulouse III University� benamara@irit.fr�Toulouse III University stdizier@irit.fr Abstract The main aim of this project is to ex- plore, develop and evaluate the contri- bution of language technologies to the development of WEBCOOP, a system that provides intelligent Cooperative re- sponses to Web queries. Such a sys- tem requires the integration of knowl- edge representation and the use of ad- vanced reasoning procedures. 1 Introduction The main aim of this project is to explore, develop and evaluate the contribution of language tech- nologies to the development of WEBCOOP, a sys- tem that provides intelligent cooperative responses to Web queries. Besides a heavy use of language (processing queries, generating responses, extract- ing knowledge from web pages), such a system requires the integration of knowledge represen- tation and the use of advanced reasoning proce- dures. Moreover, the complexity of reasoning pro- cedures must be kept reasonable in order to opti- mise tractability, efficiency, and re-usability. In a first stage, the project is developed on a relatively limited domain that includes a number of aspects of tourism (accommodation and trans- portation, which have very different characteris- tics on the web). In a second stage, we want to evaluate the re-usability of our techniques to other domains analysing where are the difficulties, what are the costs, what is domain specific and what can be shared. A number of cooperative systems were de- signed for databases in the past such as COBASE (Minock and Chu, 96) and CARMIN (Chakravarthy et al, 90). Most of the efforts were concentrated on fundamental reasoning pro- cedures, while very little attention was paid to question analysis and to NL response generation. Our challenge is to integrate reasoning procedures with real-life data extracted from web pages and to produce web style cooperative NL responses of a reasonable quality that reflect the accuracy of the reasoning procedures. A major feature is the in- tegration of a real cooperative know-how compo- nent that goes beyond the mere recognition of a user misconception. In the following sections, we briefly present the main aspects of our system focussing on question classification and knowledge representation. Co- operative response elaboration and response gen- eration are presented in (Benamara and Saint- Dizier, 03). 2 What is a Cooperative Response ? Cooperative answering systems are typically able to provide general, descriptive answers along with explanations about their answers. (Grice, 75) maxims of conversation namely the quality, quan- tity, relation and style maxims are frequently used as a basis for designing cooperative answering sys- tems. To address these behaviours, specific coop- erative techniques have been developed to iden- tify and to explain false presuppositions or var- ious types of misunderstandings found in ques- tions. Relaxation of constraints in the question 63 occur when the system cannot find any response. Finally, intentional responses may be provided in- stead of a large set of extensional answers. An overview of these aspects is given in (Gaasterland et …

Content truncated to 500 words