搜索引擎日志中“N+V+N”和“V+N+N”型短语功能类别识别

THE "N+V+N" AND "V+N+N" STRUCTURE PHRASE FUNCTION RECOGNITION IN SEARCH ENGINE QUERY LOGS

采用支持向量机(SVM)方法实现搜索引擎日志中"N+V+N"、"V+N+N"型短语功能类别识别。通过选取不同特征,构建多特征模板,实现对"N+V+N"、"V+N+N"型短语中名词短语、动词短语、主谓短语三种功能短语的自动识别,并且针对不同词性标注集对实验结果是否有影响进行了实验。实验结果显示,SVM在搜索引擎日志短语识别中有很高的识别率。

This paper proposes to use Support Vector Machine（SVM） model to recognize the function category of ＂N＋V＋N＂ and ＂V＋N＋N＂ structure phrase in search engine logs.By selecting different features,constructs the multi-feature templates to automatically recognize noun phrases,verb phrases and subject-predicate phrases three kinds of function phrases in the structure of ＂N＋V＋N＂ and ＂V＋N＋N＂ and has carried on the experiment to test the different part-of-speech tagging to experimental result＇s influence.The experimental result showed that SVM has the very high recognition rate in the search engine diary phrase recognition.