Study On Classification Of Domain-oriented User Exploration Process For Exploratory Search

Exploratory search is a specialization of information exploration which represents the activities carried out by searchers in either case:searchers are unfamiliar with the domain of their goal; search tasks themselves are relatively complex; index of information are insufficient. In the study of exploratory search, available supporting methods are proposed to meet the users'different exploratory search needs. These methods are in essence targeted at different types of user exploration processes. Therefore, better support for users to complete the exploration process could be provided by judging the types of user exploration processes and selecting appropriate supporting methods for users before applying these methods.Based on such a fact, this thesis studies the classification of user exploration processes. As a basis for modeling and classification of user exploration processes models, this thesis how to extract classification features for exploration processes. In order to formalize user exploration processes, a method which models user exploration processes based on a tree model is proposed. Based on this model, this thesis studies the classification of tree models of user exploration processes, which classifies user exploration processes by dealing with feature information of user exploration processes.In details, various feature measures for user exploration process classification systems have been considered, such as the query word feature, the search result feature, the domain feature of queries, the domain feature of search results, the query length feature, the session length feature, etc. Based on such works, this thesis utilizes a tree model to model user exploration processes. Nodes in the tree model are used to represent user exploration behaviors which contain features, and the overall structure of a tree model is used to describe an explore process. In the process of classifying users explore processes, the design of the classification function leverages exploratory information that the tree model of user exploration processes conveys. On the basis of the classification of single exploratory behaviors, degrees of relevancies between exploratory behaviors and the classifications according to the hierarchy of tree model are accumulated to determine the final classification.In the experimental stage, this thesis compares the proposed method with two baseline methods on the precision rate, recall rate and F1 rate. The experimental results show that our proposed method classifies user exploration processes more accurately, so as to better support for exploratory search.