An Auto-adaptive Query Recommendation Model Based Users' Search Satisfaction

Query recommendation(suggestion)has become an important function of web search engines,through which a user can effectively input a query that properly represent his or her information need.At the same time,Search engines can return more accurate information to meet user needs and improve the user search experience.Most existing models recommend queries by mining the user logs,knowledge base or document collections.Those works usually consider the relationships between submitted query and candidate suggestions,which neglect the user's satisfaction of current search.Those models also does not consider the importance of query suggestion to users as well as how user's satisfaction degree to current search results will influence the selection of query suggestions with respect to different queries.For the mentioned three questions,the paper firstly explores the relationship between user satisfaction and user's interest for query recommendation,as well as the relationship between user satisfaction and the novelty of the user selected recommendations.Secondly,we propose an auto-adaptive Query Recommendation Model based on user's satisfaction degree.The model can recommend novel or relevant queries to user depend on the user's current satisfaction degree.User experiments not only show a tight correlation between user satisfaction degree and user's interest to query recommendation but also show a kind of relationship between user satisfaction degree and the novelty of the user selected recommendations with respect to different queries.The experiment in the query logs based method illustrates that our model outperforms the query flow graph model.The main research work of this paper is as follows:First,introducing the research of information retrieval(IR),then we overview and summarize the query recommend model based on the different data set and the structure of model.What's more,we introduce the definition and research development of user's satisfaction degree and Query Classification.Finally,we briefly introduce the eye tracking technology applied in the field of information retrieval.Second,the user experiment of query recommendation includes two parts.The first part is an online survey through questionnaire,which records the user's education,frequency of using a search engine and search habits.It aims at exploring whether or not the users are satisfied with current search results.The second part is a user experience by using eye tracker.The experimental platform records the user satisfaction degree of current search results,the clicked queries and the users' eye movement data.Based on those acquired data,we further explore the relationship between user satisfaction degree and user's interest to query recommendation as well as the relationship between user satisfaction and the novelty of the user selected recommendations.Third,we construct an Auto-Adaptive Query Recommendation Model based on the user's satisfaction degree.We innovatively propose the model based on Query-Flow Graph that considers the influence of user's current satisfaction degree.When user are satisfied with current results,the model will recommend more novel queries,while an unsatisfied status leads to recommending more relevant queries.The performance of our model is superior to QFG,and our model has the ability in improving the user experience.