Research On Synonyms Supporting Web Tag Recommendation

With the development of web2.0, social tagging system is becoming popular. The application of tag has attracted more and more attention and useage of people. The tags are the words or phrases which describe resources, the user can label resources at liberty with no limited by existing classification system. Tag is so simple and easy to use that has been widely used. At same time the unrestricted way of use also bring problems. Because of the different understanding of vocabulary and information resources, different users are less likely to use the same tag to label the same or similar resources, so a lot of tags with similar or same means appeared in tag system. These labels increase the redundancy of the system and reduce the quality of the tag. When use the tag to search resources, if synonymous tag cannot be effectively identified, a lot of resources cannot be returned to the user which reduce the recall rate. User’s need and the tag recommended are difficult to reach a consensus, which reduce recommend quality in tag recommendation system.In view of the above questions, through the In-depth research on the phenomenon of synonymous tag, this thesis put forward a concept of synonymous tag and tag identification method based on the semantic. Through analyze the data of the tag, the causes and use features of synonymous tag are analyzed in detail. According to the closely relation of resources and semantics, this thesis puts forward a new method of finding tag semantic with the resource semantic, and recognize the synonymous tag on this basis. In the web resources of text, web text semantic is found by using the method of the probability of latent semantic. The semantics of the text is expressed as a vector of text key words. The relationship between web text and tag is analyzed by the using of the association rules mining method, and use text keywords to represent the tag semantics. A concept of tag similarity is defined, which can be used to recognize the synonymous tag. Tag similarity can be calculated by using the literally characteristics of the tag and tag semantics. The tag can be regard as the same meaning when the tag similarity is greater than a certain threshold.In order to reflecting the positive function of synonymous tag in socialization tag, the method of tag recommend based on synonymous tag is designed and implemented. This method can be used into tag recommend based on the contents of the web text. Combined with the synonymous tag, this method can produce a tag which have a complete and closely related with the web text to recommend. The result can be helpful to resource description and resource search.