问题分类中基于句法和语义信息的特征选择

问题分类是问答系统中一个非常重要的子模块,其关键在于问题的特征选择。考虑了问题的句法信息和语义信息,提出了一种利用问题疑问词、依存关系、主要动词、中心名词和名词的最高上位词作为特征进行分类的新方法。实验中,采用k-最邻近和朴素贝叶斯两种分类算法对该方法进行测试,结果表明了该方法具有较好的分类效果。在自定义的分类体系上,分别达到了82.2%和83.7%的分类精度,性能高于基于bag-of-words的特征选择方法。     

汉语中方位参考点恢复研究

方位参考点恢复是自然语言空间语义理解中十分重要问题 .方位参考点恢复是在篇章中找方位词的参考点并补充上,得到完整的空间表达式 .目前,自然语言处理技术大多面向句子级,导致省略参考点空间表达式独立出现,使空间语义理解困难 .方位参考点恢复无疑可以解决类似问题 .在此提出基于有限知识的方位参考点恢复方法 .在句法分析基础上,以知网为常识库,结合有限知识识别空间表达式以及恢复方位参考点 .实验结果表明该方法比较令人满意 .     

Hyponym extraction from the web by bootstrapping

This paper proposes an effective method to automatically extract hyponym from the Web for Chinese. The method extracts hyponyms for a given hypernym through weak supervision in two stages: the first stage is submitting a hypernym and a seed hyponym as a query to Web search engine, and automatically extracting hyponyms matching with a Chinese doubly anchored hyponymy pattern from the Web by bootstrapping. In order to reduce noise data in bootstrapping extraction, we propose a set of filtering rules to ensure matching of the proper hypernym in the extracted sentence. The second stage is ranking all the extracted candidate hyponyms by an integrated ranking algorithm which takes into account measures both of linkage frequency between coordinate hyponyms and of semantic similarity between the hypernym and candidate hyponym based on co‐occurrence statistics. © 2011 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.