Research On Social Tags Recommendation Techniques Based On Content

With the rapid development of the Internet, more and more people begin to pay attentionto social tagging which is a flexible and efficient method for classification. How to retrievetags from the huge tag library is becoming a hot topic to research. Recommendation systembased on social tags can ease users’ tagging activity, because users must tag resourcesmanually if there is no recommendation system. Frequently, it is difficult to determine thatwhich tag is more suitable for describing resources.Recommendation system based on socialtags can recommend tags automatically which are relevant to content of the resource orinterest of the user. It reduces users’ time for tagging and improves users’ experience. It issignificance that recommendation system based on social tags helps users to tag resourcesefficiently.Firstly, this paper introduces the existing systems of social tags and its recommendationprinciples used in Web2.0. Secondly, this paper summarizes the existing techniques about tagrecommendation, and analyses their merits and demerits. We find the question that only twodimensions “resource-user” are considered in most techniques for tag recommendation. Butthere are three dimensions “resource-user-tag” in recommendation system based on socialtags. This paper analyzes the overall interactive information of “resource-user-tag” andrecommends the most appropriate tags for users which can play better than other techniquesfor tag recommendation.To analyze the web pages, tags and users, this paper proposes a new method of social tagrecommendation based on content-Feature Vote Tagging(FVT). In order to study the qualityof FVT, this paper also analyses two other simple recommendation methods-recommendationbased on keyword and TF-ISTF. At last, this paper uses several kinds of evaluation methodsto assess the return results of methods. The result of experiment shows that the method whichpresented by this paper can satisfy the expect of the user for the recommendation results andresults are more precise and more suitable for users.