%A Guan Xiaoda,Lv Xueqiang,Li Zhuo,Zheng Luexing, %T Chinese Organization Name Recognition in User Query Log %0 Journal Article %D 2014 %J Data Analysis and Knowledge Discovery %R 10.11925/infotech.1003-3513.2014.01.11 %P 72-78 %V 30 %N 1 %U {https://manu44.magtech.com.cn/Jwk_infotech_wk3/CN/abstract/article_3845.shtml} %8 2014-01-25 %X [Objective] To solve the problems of query log annotated data shortage and information asymmetry in user query log organization name recognition. [Methods] The paper proposes an automatic method to create training data,which abates the insufficient of user query log annotated data. The authors cite the adhesion features and constructed CRF model to recognize organization names by integrating context information. [Results] Experiments on Sogou user query log show that precision rate can reach 72.80%,recall rate can reach 86.73% and F-measure can reach 79.16%. The method improves F-measure by 30% comparing with the traditional organization name recognition method. [Limitations] The model error using auto-created training set will be greater than standard annotated user query log data.The scale of organization name set will affect the completeness of the model’s context knowledge. [Conclusions] Experiment results demonstrate that the method is effective.