摘要
针对话题追踪与检测多停留在二维空间的平面集合操作,忽略了事件主题及其直接相关事件之间可能存在一定的层次关系这一问题,通过改进蚁群聚类算法中的相似度度量方法以及状态转换函数改进现有蚁群聚类算法,并利用改进的蚁群聚类算法实现新闻话题的子话题自动划分。结果表明,改进的算法能够具有较高的子话题划分识别能力。
In order to solve the problem that most of the researches on topic detection and tracking stayed in two-dimensional spaces, and researchers ignored the event topics and their hierarchical relationships among events, the ant colony algorithm was introduced to solve the subject classification. An improved ant colony algorithm was proposed by improving the similarity calculation method as well as the state transition function of the existing ant colony algorithm, and was implied to realize automatic subtopic partition of news topics. The results show that the improved method has higher abilities of subtopic partition and recognition in the process of topic tracking.
作者
韩冰
汪波
HAN Bing WANG Bo(College of Management and Economics, Tianjin University, Tianjin 730000, Chin)
出处
《济南大学学报（自然科学版）》
CAS
北大核心
2016年第6期473-478,共6页
Journal of University of Jinan(Science and Technology)
基金
国家自然科学基金项目(71272149)
教育部人文社会科学研究基金项目(11YJA630076)
关键词
蚁群算法
话题追踪与检测
子话题划分
ant colony algorithm
topic tracking and detection
subtopic partition
