自然言語入力によるWebを情報源とした質問応答

概要

論文の詳細を見る
Recently, the question answering system using natural language has been actively researched by increasing of the demand for the web retrieving. But, it is very difficult to realize the question answering system using natural language, because the accuracy in the syntactic and semantic analysis is not enough. To realize a practical system, we propose a method for the question answering system using table structure. We assume that humans understand conversation by a simple method instead of the syntactic and semantic analysis. This method is table structure. The table structure is a kind of table that integrated the subject and predicate in a sentence. In the proposed method, the answer is obtained by comparing a table structure of question sentence and target sentences. So only morphologic analysis is performed in this method. In this report, we describe the evaluation of performance and the result of some experiments in the proposed method.
北海道工業大学の論文
2007-03-26