Automatic detection of descriptive phrases for Question Answering System:

The increasing interest in providing various information on the Web has heightened the need for a sophisticated search tool. Most existing information retrieval systems, however, merely provides documents, and this often leaves users to read a relatively large amount of full-text. The study of question answering (QA) systems, which enable people to locate the information they need directly from large free-text databases by utilising their queries, has become an important aspect of information retrieval research. The purpose of this work was to evaluate a Descriptive Phrase Detection (DPD) system that attempted to detect descriptive phrases from a free-text database. A descriptive phrase was a phrase that explained or described a word/noun phrase. Those detected phrases were expected to be the candidates, that could answer a particular class of question in a QA system. Those questions could be ’What is sushi?’, ’Who is Bruce Brown?’, ’What job does Steve Jobs do?’ or ’What does ISDN stand for?’ This system employed only simple pattern matching for detection and term frequency for ranking in order to achieve topic domain independence and to allow the use of free-text as the information source. As a result of the experiment with 57 queries, the system succeeded in detecting a descriptive phrase in more than 70 percent of the queries, and was able to rank the phrases of 60 percent in the top 5, and 70 percent and 80 percent in the top 10 and 20 respectively. These findings suggest that a system which employs only simple pattern matching and term frequency ranking, has the potential to provide the descriptive information of a word/noun phrase, with the use of a free-text database.