Design And Implementation Of User Satisfaction Sysytem Based On Search Engine Logs

The rapid development of Internet has become a crucial way for the public to obtain information from the data bank.From the first generation of classified catalogue retrieval,the technical core was transferred and upgraded constantly.In the age of the user,user's satisfaction is regarded as one of the essential criteria of search engine to identify the quality risk.Any commercial operation requires some kind of “bias” from users to control the direction.Therefore,evaluate user satisfaction effectively and mining Bad Case is important,it is a vital safeguard for search engine to realize system optimization,algorithm improvement and routine maintenance.The assessment system(User Satisfaction Assessment System,USAS)is designed to meet the evaluation needs of search engine,it is for user satisfaction and Bad Case mining based on theoretical framework and relevant evaluation study.The main work is:(1)The overall design scheme and realization route are proposed,and the study objective is to think of user experience not in terms of what the answer they get,but how they are feeling in this process.(2)To speed up log processing,the distributed processing technology and data cleaning method is combined in the system architecture.For collecting the real perception of users,trajectory retrieval technology is used to restore and label the click information.(3)After analysis the commonalities and special characteristics of user behavior,a strategy of feature extraction is proposed.It enhances the descriptive power.(4)The strategy of using neural network to build satisfaction evaluation model and Bad Case mining model is proposed,which solves the problem of non-linear correlation between user's browsing behavior and user's satisfaction,it improves the accuracy and reliability of USAS system.(5)The experiment results of analyzing the user satisfaction based on search logs indicate that traditional way is incomplete for obtain the real feelings of users,and modeling the Bad Cases show that the negative factor may be relevant,which means that one solution could solving a series of questions.Those analysis data is generated by automatically,and it provided a reliable evidence of technology optimization for search engine.