摘要
As increasing amount of social data on today’s social web systems,user-generated contents are not only getting richer, but also frequently interconnected with users and other objects in various ways. Social data provides a perfect platform for personalized Web search. However, it is confronted with a great challenge named vocabulary mismatch problem. To overcome this problem,previous research has proposed many effective approaches utilizing social query expansion based on co-occurrence statistics, tag-tag relationships and semantic matching etc. Most of them focus on the statistical relationships between words/terms ignoring their truer semantics. In this paper, we propose a novel generative model which uses word embeddings to cluster words to enhance the latent topic model. Instead of just relying on the statistical relationships of words, the approach tries to take into consideration of semantic information of context words and word clusters to construct user models for personalized query expansion. Experimental results on a large public social dataset show that the proposed method is more effective than other state-of-the-art baselines.
出处
《国际计算机前沿大会会议论文集》
2017年第2期15-17,共3页
International Conference of Pioneering Computer Scientists, Engineers and Educators(ICPCSEE)