Research On Ranking Answer For Dataspace

The past few years with the development of science and technology, more andmore types of document should be kept. Therefore, a new type of data managementsystem, DataSpace which can store all types of documents and also can integratebetween them is proposed. DataSpace Support Platform (DSSP) is similar toDatabase Management System (DBMS), provides the facilities of the searchinformation. The quality of a search engine can be judged from the accuracy ofreturned information.This thesis studies how to rank search result in DataSpace effectively. In the past,ranking search result is depending on the index method of the storage device.However, this thesis focuses on using implicit feedback as a factor to get userbehavior. There are many factor has been used for ranking with implicit feedback. Inthis thesis, we propose a rank algorithm based on User Activity Record (UAR)analysis and add Semantic Similarity (SS) to find other synonym word of the searchkeyword.UAR is an attribute which adapts activity theory to determine the importance ofa document for the user. This method records all user activity information, then usedan algorithm analyzes the user’s activity record to find the relationship amongst threefactors: duration for user opens a document, elapse-time for the next re-opening adocument, frequency opening a document within a specified period of time; and theimportance of the document. The algorithm gets the user activity information recordsas data source and through analysis the duration of activity and the frequency ofactivity calculate the rank of the activity.Then in finding the relevance between query and document, we use SemanticSimilarity (SS) which counts the similarity between two words. Firstly, get synonymwords using WordNet database, then using an algorithm to find the SS betweenkeyword and other synonym words. Then add the highest similarity degree word as asearch keyword in DataSpace.This paper regards ranking as the factor to get better feedback document for theuser query. Preliminary experiments show that the USR and SS can give better list of document feedback and provide more effective and more efficient service for user insearching the DataSpace.