Dialogue Helpsystem based on Flexible Matching of User Query with Natural Language Knowledge Base

Abstract

This paper describes a dialog helpsystem which advises users in using computer facilities and software applications provided by the Center for Information and Multimedia Studies, Kyoto University. The system employs a knowledge base written in natural language and retrieves a proper knowledge unit by flexible matching of user query with the knowledge base. The system is running since July 1999, received about 2,000 queries for the first seven months, and answered about 40~ of them satisfactory. 1 I n t r o d u c t i o n One of the ultimate goals of Natural Language Processing is to realize a dialogue system which can communicate with human beings in a natural way (Wilks, 1999). However, no effective real world dialogue application exists so far, not only in spoken language for which speech recognition is still a big obstacle, but also in typing language. The most serious problem is knowledge. It is obvious that without sufficient knowledge a dialogue system cannot talk with people satisfactorily. Classical dialogue systems like UC (Wilensky et al., 1984) utilized a formal language to represent knowledge, which requires the heavy cost of construction and maintenance and makes the scaling up quite difficult. In contrast, along with the improvement of NLP, research activities which utilize natural language text as a knowledge base become popular, such as START (Katz, 1990), FAQ Finder (Cooper, 1996), and QA Track in TREC 8 (NIST and DARPA, 2000). These systems, however, basically produce one time response, and do not have a conversation with users. This paper proposes a dialogue helpsystem in which natural language knowledge base is not only used for one time response, but also for conducting a conversation. To put it concretely, the system can • ask the user back if there is an unknown word in the user utterance, • interpret the user utterance contextually, and • ask the user back when the user utterance is not clear. In order to achieve these functions, a mild restriction is imposed on the knowledge base; it is compiled in a dictionary-like structure (entry and its description) and the description is written in an if-then format (though several expressions can be used to show ifthen relation). The dialogue helpsystem reported in this paper advises faculties and students in using the computer facilities and software applications of a university. The computer facilities of Kyoto University, about 1,200 computers (PCs and WSs) in several rooms and modem connections are administrated by the Center for Information and Multimedia Studies (CIMS; http://www.media.kyotou.ac.jp/). The situation at CIMS is that many people are using computers, teaching

Extracted Key Phrases

5 Figures and Tables

Cite this paper

@inproceedings{Kurohashi2000DialogueHB, title={Dialogue Helpsystem based on Flexible Matching of User Query with Natural Language Knowledge Base}, author={Sadao Kurohashi and Wataru Higasa}, booktitle={SIGDIAL Workshop}, year={2000} }