关键词: 热点话题, 主题词, 微博文本, 联合排序, 热度序列

Abstract: Micro-blog hot topic detection plays an important role in public opinion analysis and opinion mining. In   order to reduce the impact of data sparsity on topic detection, this paper proposes an approach for micro-blog hot topic detection based on heat co-ranking, builds a unified model framework to organically integrate all relationships between micro-blog texts and topic keywords. The authority of micro-blog user and the time-related characteristics of topic keywords are simultaneously considered, and the heat of micro-blog texts and topic keywords gets mutual reinforcement and co-ranking. Topic keywords in hot sequence are clustered by using the combination support confidence as a threshold. The experimental results show that the proposed method has high accuracy for hot keywords extraction and hot topic detection, can effectively discover micro-blog hot topics in a specific period.

Key words: hot topic, topic keywords, micro-blog text, co-ranking, heat sequence