Read/Search this Article
Abstract
人物に関する質問応答を想定し、機械学習に基づき、テキストから人物の説明記述を精度よく抽出する一手法を提案する。この手法は次の二段階の洗練化に基づく。まず、テキスト中の各パッセージについて、人物の説明記述に関連するか否かの判断を行う。次いで、抽出されたパッセージについて、説明記述の範囲を詳細に決定する。評価実験によれば、文列に対する系列ラベリングに基づく一段階の抽出手法と比較して、提案手法の方が高精度であった。
We propose a method of extracting descriptions about a person from text by machine larning. It is intended to be utilized in the question-answering about person. This method is based on a multistage system that consists of two types of classifiers. The earlier classifier judges whether each passage includes description about a person or not. The later one decided the exact extent of the description in passages extracted by the earlier classifier. The experimental result shows that the proposed method is more accurate than the one-staged chunking method based on the labeling on a sentence sequence.
Journal
- IEICE technical report. Natural language understanding and models of communication [List of Volumes]
-
IEICE technical report. Natural language understanding and models of communication 108(141), 79-84, 2008-07-10 [Table of Contents]
The Institute of Electronics, Information and Communication Engineers