摘要
通过利用Markov网络团的方法来对查询意图识别.首先从人工标注搜狗查询日志中约2 250个查询作为测试数据,采用搜狗提供的分类语料(共10类)来建立Markov网络,用建立的Markov网络来对查询进行扩展,得到相关的返回结果列表,运用在分类语料训练好的分类器来对返回结果进行分类,从而完成对查询意图识别的过程.实验中采用的评价指标是11_avg和3_avg,实验结果表明该方法能够有效地提高检索效率.
A new method about query intent classification is proposed. Making use of manually labeled queries form Sogou＇s query log （about 2 250） as training data, and use the ten classes of data to construct the Markov net- work. So we can effectively get information of the queries. After this process research the queries in this data. Returning the relevance results of the queries, and classifying these results according to the classifier trained by the ten classes data. At last, the queries intent is predicted. In the experiments use the 11_avg and 3_avg as assessment process. Experiment results demonstrate that the algorithm presents some advantages compared with other methods.
出处
《江西师范大学学报（自然科学版）》
CAS
北大核心
2012年第4期383-387,394,共6页
Journal of Jiangxi Normal University(Natural Science Edition)
基金
国家自然科学基金(60963014)
江西省自然科学基金(20114BAB201037)资助项目
