基于答案辅助的半监督问题分类方法.PDFVIP

文章编号：1003-0077 （2011）00-0000-00 基于答案辅助的半监督问题分类方法 张栋 李寿山 周国栋 苏州大学自然语言处理实验室 苏州 215006 E-mail: d.zhang.suda@, {lishoushan, gdzhou}@ 摘 要：问题分类旨在对问题的类型进行自动分类，该任务是问答系统研究的一项基本任务。本文提出了 一种基于答案辅助的半监督问题分类方法。首先，将答案特征结合问题特征一起实现样本表示；然后，利 用标签传播方法对已标注问题训练分类器，自动标注未标注问题的类别；最后，将初始标注的问题和自动 标注的问题合并作为训练样本，利用最大熵模型对问题的测试文本进行分类。实验结果表明，本文提出的 基于答案辅助的半监督分类方法能够充分利用未标注样本提升性能，明显优于其他的基准方法。 关键词：问答系统；问题分类；答案辅助；半监督分类；标签传播 中图分类号：TP391 文献标识码：A Semi-supervised Question Classification with Answers Dong Zhang, Shoushan Li, Guodong Zhou Natural Language Processing Lab of Soochow University, Suzhou, 215006 E-mail: d.zhang.suda@, {lishoushan, gdzhou}@ Abstract ：Question classification aims at classifying the types of questions automatically, and this task is a basic task of question answering system. This paper proposes a method of semi-supervised question classification with answers. Firstly, answer features bind question features to realize samples said; Secondly, a question classifier will be trained on labeled questions using label propagation algorithm to annotate the category of unlabeled questions automatically. Finally, the questions of initial annotation and automatic annotation are merged with each other as training samples, using maximum entropy model to classify the testing samples. The experimental result demonstrates that the method of semi-supervised question classification with answers in this paper can make full use of the unlabeled samples to improve the performance, and is better than other benchmark methods. Keywords: Question answering system; Question classification; Answer aiding; Semi-supervised classification; label propagation 1 引言 问答系统能够为用户提出的自然语言问题提供一个简明、准确的答案，越来越受到人们 的关注。现有的问答系统主要包括三个模块：问题分析、信息检索和答案抽取。问答系统为 了能够正确回答用户所提出的问题，首先需要对问题进行分析，知道用户想要寻找什么信息。