Distinction Between Consensus Tags And Non-consensus Tags Based On User Intention

With the rapid development of Web2.0, tag begins to be concerned and used by more and more people as one of the typical applications. tag is a short word or phase related to a web resource. Users can categorize resources by putting several tags on resources. The feature that tag can be managed freely by user makes use of user’s knowledge well and bring the tag information processing a big challenge. How to organize resources properly using tags becomes a hot research currently. Existing related research of tags based on the consensus of tag labeling, mainly focus on the relationship between tags, but rarely considers the relationship between tags and resources.Aiming at the problems above, this thesis proposes the consepts of consensus tags and non-consensus tags, and proposes distinction of consensus tags and non-consensus tags by deeply analysis of tag consensus. Firstly, this thesis analyzes the behaviour of tag labeling, deeply analyzes the causes of consensus tags and non-consensus tags and the evolution of tag consensus and introduces the consepts of consensus tags and non-consensus tags on this basis, then gives an overall description of framework of distinction. The theory of distincion of consensus tags and non-consensus tagsis is as follows:select the co-occurrence words of the tag from the blog post, then merge all the co-occurrence words of the tag to form the associated words set of the tag; Determine whether the tag is a consensus tag according to the set. This thesis proposes the selecting method of co-occurrence words based on KeyGraph and the determining method of consensus tags based on Subjective Bayesian Method. The key of the selecting method of co-occurrence words based on KeyGraph is selecting words, which are closely associated with a particular tag, from the blog post in order to reflect the meaning of the tag on the blog post. The determining method of consensus tags based on Subjective Bayesian Method merges all the co-occurrence words of a tag to form the associated words set of the tag which reflects the general laws of tag labeling. The method takes the associated words set as the input data, calculating the the probability for the consensus of the tag by Subjective Bayesian Method in order to determine whether a tag is a consensus tag. Finally, this thesis introduces two application of consensus tags:consensus tag supplement and division of associated words set. consensus tag supplement is used to supplement blog post tags, division of associated words set is used to analyze the polysemy of tags. The experimental results is used to prove the effectiveness of the algorithm that this thesis proposes.