A Study Of Readability Based Information Retrieval Model

Information retrieval is the activity of obtaining information resources relevant to an information need from a collection of information resources.With the increasing diversity of both content available from various kinds of resources and users,how to provide users with both relevant and readable document has been urgent problem.In order to solve this problem,we tried to integrate readability into information retrieval model to improve the satisfaction of users.In this paper,we proposed both ontology-based and sequential topic-based readability model for domain specific field.In the ontology-based readability model,we take both average conceptual scope and sequential dependency into consideration to measure documents' readability from both conceptual level and lexical level.We construct the domain taxonomy automatically in sequential topic-based readability model to improve its flexibility.Based on the process of reading and understanding,this model can also measure documents' readability in both topic-based and lexical level.As a subjective concept of readability,we attempted to integrate users' reading level to optimize readability measures which shed light in the interactive information retrieval direction.To verify the effectiveness of proposed readability models or measure,both user-oriented and system-oriented evaluations have been conducted.The former are used to test the consistence of model prediction and user judgments,while the latter is to find out whether the proposed models are effective in information retrieval.Results show that sequential dependency employed in the ontology-based readability model gained meaningful clues.In the sequential topic-based readability model,not the results of user evaluation show good correlation with users' judgment,but also we got significant improvement in system evaluation.While,through integrating user reading level to readability computation can improve the precision of readability prediction.