|
Social tagging is a totally user-driven data structure, which allows human users to organize any kinds of online resources (e.g., URLs, videos, documents, photos and so on) by means of words or phases which are called tags. Users choose the tags which can represent the users’ views and reflect the preference of users. The way of connecting between users and the system results in the triple structure of users, resources and tags.There is no restriction for users to take part in the social tagging system, they can classify and category their resources freely, what’s more, the users can also find the other users with the same mind.Based on it, social tagging has some strong points, such as least effort retrieval, low cognitive cost and quick adaptation to changes in words and resources, etc. However, the development of social tagging still faces to a lot of serious challenges, the core problem is the ambiguity of tags which is also the hot topic nowadays.In this paper, first of all, we made a deep research and understanding on relevant concepts, system structure, advantages and problems of social tagging. Then, we discussed and analyzed the algorithms of the two main methods in social tagging field which are based on the triple structure and semantic analysis. Finally, we proposed a novel method for tag analysis based on latent semantic analysis (LSA) which is called tag clustering based on latent semantic analysis (LSAC) which can make tags into semantic related groups.In order to verify the effectiveness and reasonable of LSAC, we used the real-life dataset which extracted from Movielens to make a series of experiments compared with the classic clustering method Kmeans. The experiment results show that our method has a better performance on the task of obtaining tags’ real meaning.
|