Tag Recommendation Based On Collaborative Filtering And Text Similarity And Search Optimization

At present, social tagging system has become the research focus, users can freely add tags for the uploaded resources in the system. Because each user's different knowledge backgrounds, which caused a problem that many tags could not describe the resource properly and even have some spelling errors. This problem hinder search this kind of resource for other users. The accuracy of tag recommended is an important part of the promotion of social tagging system.In the social tagging system, tags of users tagging can not better describe the contents of the resource, which not only affect the resource sharing and classification, but also not conducive to promoting better. In addition, using keywords are too vague when users search for resources, the traditional keyword matching method can not meet the needs of users. For this, we proposed a tag recommendation method based on collaborative filtering and text similarity and search optimization. For the webpage of users would tagging, calculating the number of users who have marked the input webpage. If marking times is more than the threshold, the collaborative filtering method is used. Search users that are similar with the current user, the tags with higher weights are recommended to the current user, else we use the improved text similarity formula to search pages, and the tags with higher weights are recommended to the current user. But the direct use of cosine similarity formula can not better solve the tagging problem. In this paper, we improve the cosine similarity formula in calculating similarity webpages, calculate the weights of each word and each webpage. Using improved cosine similarity formula can be good solve these problems when calculate similar webpages. Finally, this paper analyzes the role of tags in optimize web search. When query terms are not in a webpage, using traditional methods can not get good results. For this, the article improves the formula of search for the target webpage, using the tag social characteristics. At the same time, we consider the tag marked follow the power law distribution, taking the logarithmic approach reduce calculation errors.Experimental results show that this algorithm has effectively improved the accuracy of tags recommended,comparing with the current tags recommended algorithms.