Statistical Methods based on Semantic Similarity of Topics Related to Microblogging

摘要

Existing topic tracking methods are mostly for news and forum data, which lack of statistical methods for microblogging on relevant topics. Combined with characteristics of micro-blog information, the paper proposes a microblogging statistical methods based on semantic similarity. Firstly by building topic semantic model and then use the HowNet semantic similarity calculation of two terms, and measures the relevance of the topic and microblogging. Finally statistics method is provided on the degree of correlation. Experiments show that novel method works on the problem soundly.