A User Model Construction Method Based on Tag Disambiguation

Abstract

The invention provides a user model building method based on label disambiguation. The method includes the steps of obtaining website user labeling data from a label website, building a labeling resource information database, building a user model and a resource model, recognizing a polysemic label rp for the user model, determining a semantic item and a neighbor label set of the polysemic label rp, generating a disambiguated user model, embedding the user model to a label website background, and carrying out resource pushing according to the cosine similarity between the user model and the resource model. The polysemic label in the user model is disambiguated considering label polysemy and obstruction on accurate information recommendation from label polysemy, semanteme of the user model is clearer, misguided information recommendation caused by label ambiguity is avoided, and therefore the method provides a support for better personalized information recommendation services of multiple labeling websites.

Description

technical field [0001] The invention relates to social tagging technology, in particular to a method for constructing a user model based on tag disambiguation. Background technique [0002] With the rise of e-commerce and social networking sites such as Delicious, YouTube, Flickr, and Movielens, user information is unprecedentedly abundant, but how to provide users with more effective personalized recommendation services is increasingly becoming a challenge. Social annotation provides a new idea for the construction of user models in personalized research. Researchers have proposed a variety of recommendation algorithms, which not only improve the efficiency of recommendation, but also enrich the research in the field of personalized services. However, there are also some imperfections in social labeling, and the ambiguity of the vocabulary used in the label is one of the typical problems. In the absence of context, people are often unable to correctly understand the exact ...

AI Technical Summary

Problems solved by technology

However, in the personalized services of tagging websites such as Delicious and YouTube, whether it is the user model or the resource model, on the one hand, a large number of tags are involved in the model, so there may be multiple ambiguous tags, and if Adding additional information to the model for each label is very likely to overwhelm the original information in the model and generate new unpredictable contexts; on the other hand, the semantic information between labels in the user model is weak and cannot be passed through each other. The association between them produces clear semantics, which makes it difficult to disambiguate polysemous words

The insufficiency of such a model makes it impossible for tagged websites to recognize the correct meaning of tags such as apple and SF during the recommendation process, and to recommend irrelevant but consistent word forms to users, making it impossible to personalize tagged websites. The recommendation plays an accurate guiding role

Images