Micro-blog Recommendation Based On Word2vec Topic Extraction

Recently, faced with the problem of information overload on the micro-blog, Some personalized information and friend recommendation for micro-blog has been proposed. However, there are many researches of micro-blog is on relationships of friends and community structure, less researches is on the micro-blog content currently. In reality, users have to face all the information released by the followed. There is a great deal of information cannot attract users, and users are often unable to track and control subjects they concerned. In order to extract the topic of micro-blog and realize content-based recommendation under different themes, this article will explore in depth.We propose a recommendation framework based on the word2 vec micro-blog topic extraction in this paper. The framework can extract topic from user’s history data,then filter new micro-blog information to remove the content users are not interested in, and match the content that users are interested with the topic,at last achieve topic-based recommendation.Considering the micro-blog content is composed of short text,this paper also proposes a clustering method based on word2 vec to realize the topic extraction.At the sametime, in order to verify the accuracy of recommendation framework, we conducted a series of comparative experiments,such as comparing the general classification with the classification which consider various features of micro-blog, and comparing the clustering algorithm based on word2 vec with LDA, K-means clustering algorithm,and comparing content-based recommendations under the topic with some of the existing micro-blog recommendation algorithm. Experimental results show that the recommendation framework proposed is feasible.