摘要
[目的/意义]主题识别研究对于理清领域内的知识结构与研究热点非常重要,对领域主题进行动态识别,可以很好地帮助研究人员了解和掌握领域的发展态势及未来走向。[方法/过程]利用张量的数据结构形式,在词共现矩阵中融入时间维度,只需一次聚类便可进行动态主题的识别。[结果/结论]张量结构及非负张量分解算法为词共现频次变化视角下的动态主题识别提供一种新的方法,该方法相较于传统方法更为简单快捷,有效避免了信息的损失。
[Purpose/Significance]The research on topic recognition is very important to clarify the knowledge structure and research hotspots in the field.Dynamic identification of domain topics can help researchers understand and master the development trend and future trend of the field.[Method/Process]Using the data structure form of tensor,this paper integrated the time dimension into the word co-occurrence matrix,and only needed one clustering to identify the dynamic topic.[Result/Conclusion]Tensor structure and non-negative tensor decomposition algorithm provide a new method for dynamic topic recognition from the perspective of word co-occurrence frequency change.Compared with traditional methods,this method is simpler and faster,and effectively avoids the loss of information.
作者
席崇俊
刘文斌
丁楷
Xi Chongjun;Liu Wenbin;Ding Kai(Institute of Science and Technology Information of China,Beijing 100038)
出处
《知识管理论坛》
2022年第2期197-208,共12页
Knowledge Management Forum
关键词
关键词共现
非负矩阵分解
非负张量分解
动态主题识别
知识管理
co-occurrence
non-negative matrix factorization
non-negative tensor factorization
dynamic topic recognition
knowledge management
作者简介
席崇俊,硕士研究生,E-mail:xicj7465@163.com;刘文斌,硕士研究生;丁楷,硕士研究生。
