Based On Interest Polygon Personalized Search System

Traditional search engines return the results only depending on the keywords entered. Different users'personal requirements are not considered. Actually, many keywords searched are in two or more fields, so the search results are also in different fields. User only needs the results in his target field and wishes results in other fields be eliminated, what exactly the personalized search does.Personalized search is a new search mechanism which collects user's operations and predicts user's interest. User's interest model is generated and the similarity matches between user and search results are done to optimize the rank of results. How to generate an accurate efficient interest model, how to dynamically adjust the model like human changing his interest and how to deal the similarity match are big issues in personalized search area.Based on the issues mentioned above, a new personalized search method based on a new two-dimension geometrical model called interest polygon model is proposed. User's interest is displayed as a polygon by mapping the values in vector model into two-dimension plane. A new similarity match method which counts the overlapped area between two interest polygons is proposed for the model. The interest background which includes the interest polygon and the dynamic adjustment strategy are also researched in the paper. The interest background is served for similarity match and it is composed of fields'distributing and valued areas. With combining explicit feedback and implicit feedback, the dynamic adjustment strategy can factually simulate human to change their interest. The interest polygon and dynamic adjustment strategies solve the key problem of generating and adjusting interest model.At the end of paper, a personalized search system which uses interest polygon model is realized. The experiment result shows that the personalized search method can effectively improve the search result ranking according to user's interest model and fulfills user's personal requirement. The echo time and rank optimization rate are also fine and acceptable.