<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0507">
  <Title>A Practical QA System in Restricted Domains</Title>
  <Section position="2" start_page="0" end_page="1" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> During the last decade, automatic question-answering has become an interesting research field and resulted in a significant improvement in its performance, which has been largely driven by the TREC (Text REtrieval Conference) QA Track (Voorhees, 2004). The best of the systems in the QA Track is able to answer questions correctly 70% of the time (Light et al., 2003). The 70% of accuracy is, of course, high enough to surprise the researchers of this field, but, on the other hand, the accuracy is not enough to satisfy the normal users in the real world, who expect more precise answers.</Paragraph>
    <Paragraph position="1"> The difficulty of constructing open-domain knowledge base is one reason for the difficulties of open-domain question answering. Since question answering requires understanding of natural language text, the QA system requires much linguistic and common knowledge for answering correctly. The simplest approach to improve the accuracy of a question answering system might be restricting the domain it covers. By restricting the question domain, the size of knowledge base to build becomes smaller.</Paragraph>
    <Paragraph position="2"> This paper describes our restricted domain question answering system for an agent robot in home environment. One of the roles of the home agent robot is to be able to answer the practical questions such as weather information, stock quote, TV broadcasting schedule, traffic information etc. via a speech interface. The agent should provide high-precision answers, otherwise the users will not trust the entire functions of the home agent robot, which includes not only the ability of question answering but also the speech interface for controlling home appliances. That means no answer is preferred to a wrong answer and the primary concern in our research is improving the precision of the question answering system.</Paragraph>
    <Paragraph position="3"> In this paper, we present a question answering system which is restricted to answer only to the questions on weather forecasts  , and provide some experimental results of the restricted QA system.</Paragraph>
    <Paragraph position="4"> To achieve the high accuracy, the QA system processes the semi-structured text data on the Internet and store it in the form of relational database. The domain specific hand-coded ontology containing weather terms and cities is manually built for the question analysis and the inference process.</Paragraph>
    <Paragraph position="5"> The remainder of the paper is organized as follows. Section 2 describes the overall architecture of the QA system. Section 3 describes the practical QA system. Section 4 evaluates the system and reports the limitation of the QA system. Section 5 compares our system with other QA systems. Section 6 concludes with some directions for future work.</Paragraph>
  </Section>
class="xml-element"></Paper>