用户查询意图切分的研究

Study on Segmentation of User's Query Intents

用户查询意图是指用户在构建查询时,希望搜索引擎能够返回的信息.如果搜索引擎可以判断用户当前查询与之前输入的查询是否属于同一查询意图,那么可以为用户提供更适当的查询建议、查询扩展或者个性化检索服务等.该文提出了基于点击相似度切分用户查询意图,在决策树模型和CRF模型上都取得了一定的提升.利用用户点击信息可以提高查询意图切分的效果,引入用户点击信息后,基于决策树的方法,F值提高1%,基于CRF模型的F值提高1.4%.

Query intent denotes the specific information that user wants to look for when he sub- mit a query to a search engine. If search engine can judge whether the intent of current query is same as that of previous query, they would refine current query and offer users more accurate results and services, such as query suggestion, query expansion or personalized information retrieval. This paper proposed click data-based similarity, and this similarity is effective with decision tree model and CRF model. Users＇ click data is helpful for this task, with users＇ click data, F-measure is improved by 1% and 1.4% for decision tree-based method and CRF-based method respectively.