Tag Recommendation Research Based On Tag-Topic Model

Social tagging system is a mechanism for the users to organize, manage and share resources. Users can be allowed to add any words to the resources, which the words composed the social labeling. It can be learn easily and operate simply by the network users. With the development of Web2.0, it becomes the one of the currently most popular applications.Some questions exist in the social labels, such as the noise labels, the use of non-specification, the low-rate utilization and the sparse distribution. In order to improve the quality of the social labels, tag recommendation technology emerged as the times require and has been widely concerned in recent years and it has become a hot research field of information retrieval.This paper focuses on the tag recommendation technology in the book social tagging system. The work shows as follows:Firstly, it proposed a tag recommendation method based on tag-topic model. The new text is presented the set of words, while using the trained tag-topic model to calculate the probability which the tags generated words. And using the probability which the tags generated words to calculate the probability between each word of the set and all the tag, finally, the tags which related to the new text are found. On this basis, adding a TF-IDF value auxiliary to improve the final tag recommendation result. The experiment results showed that it can improve the recommended effectively in the advantage of the implicit topic of this coarse-grained level by increasing the description of the word characteristics of this fine-grained level, and the comparative experiment results verify the desirability of this method.Secondly, a tag recommendation system based on a book database completed. The system contains three modules:the tag recommendation module, the system obtains a introduction of a book, which the user inputted and processes it, and recommends ten tags by using the tag-topic model which had trained before; the book browsing and retrieval module, the page shows the most frequently used labels counted from the database, click on one tag to get a page, which shows a list of all the books in the database that labeling with the tag, as well as get the same page by retrieve the same tag; the book adding module, the user input the information of the book as the required format, and the system recommend tags for the user to choose by using the information of inputted book, finally the information of the book and the chosen tags are written to the database.