关键词: 网络热词, PAT-Tree, 互信息, 中文字符串, 候选短语

Abstract: This paper proposes a candidate phrase extraction methods based on PAT-Tree. By modifying the PAT-Tree data structure, the paper makes it suitable for the Chinese string of variable length, then uses mutual information to assess the candidates. Combined with news text’s features and characteristics of network hot words, the paper uses a forward filtering method to filter the candidates. Compared with other similar algorithms, our algorithm does not need complex language rules and evaluate formula. The experimental results show that our algorithm is proper and efficient.

Key words: network hot word, PAT-Tree, mutual information, Chinese string, candidate phrase