Research And Implementation Of Search And Recommendation System

Currently,Web resources become more and more rich in content and widely spread in the web,at the same time become the most important source of information.It is useful to help users find information accurately,thus it has become a hot research topic in recent years.This Paper is on background of research work in Pwaltz system which has recommendation and search function,for this reason we design and develop a Web search engine that can provide recommendation function.This paper gives an overview of IR technologies and some recommendation theory.Based on the traditional information retrieval and recommendation methods,we do some research to improve the search and recommendation function.The main research work includes following aspect:(1) Research work on search algorithm based on vector space model.Traditional search engines are lack of semantic understanding,thus we propose a search algorithm based on vector space model.The algorithm uses statistical method to establish a "document-Latent Semantic-word" relationship between document and word in the probability distribution,and to use this relationship to retrieve information.When the user inputs a query on the Pwaltz search engine,the system first use the VSM model to retrieve document that contains the keyword,and then use PLSA model to find and order the document with the score that most relevant with input latent semantic.Experiments show that improvements can be obtained by using the proposed approach.(2) Research work on recommendation algorithm in Pwaltz system.We formalize the recommendation as a ranking problem,and propose a Random Walk-based machine learning algorithm to adjust parameters over the heterogeneous resources network.By using machine learning method to identify different types of edge,the algorithm provides users with information according to its importance and similarity which are calculated offiine, Experiments show that improvements can be obtained by using our proposed approach.(3)Construction of a system with support of the search and recommendation function.According to the requirements in Pwaltz system,by using JAVA Programming we realize MVC-based retrieval system.The system optimizes it's search and recommendation features based on Lucene(full-text search technology),improves search algorithm based on the vector space model,and adjusts parameters over Heterogeneous Network.Through presentation of the system,we provide the procedures from establishment of the index of document to the accurate information retrieval and multi-type resources recommendation.To sum up,the method proposed by this paper has great advantage:it is simpler and can be easily understood by software engineering staff.It does not only deserve a closer study, but also has good engineering application value.