Abstract: This paper analyzes the formation of the Chinese abbreviations, defines the morphology similarity between two words, and proposes the story similarity computation method based on the longest string matching. It explores the usage of this similarity computation method in the Chinese report link recognition system. Experimental results show this method performes well, reduces the normalized detection cost by 12.96%, and greatly improves the performance of the story link recognition system.

Key words: report link recognition, topic detection and tracking, abbreviation, normalized detection cost, similarity computation method

摘要： 分析中文缩略语的构词方式，定义2个词之间的词形相似度，提出一种基于最长字符串匹配的相似度计算方法，探讨该方法在中文报道关系识别系统中的应用。实验结果表明，该相似度计算方法能够改善中文报道关系识别系统的性能，使系统的归一化检测开销降低12.96%，取得较好的识别效果。

关键词: 报道关系识别, 话题检测与跟踪, 缩略语, 归一化检测开销, 相似度计算方法

CLC Number: 

• N945