Hits:

Indexed by:期刊论文

Date of Publication:2015-07-15

Journal:Journal of Computational Information Systems

Included Journals:EI、Scopus

Volume:11

Issue:14

Page Number:5031-5038

ISSN No.:15539105

Abstract:Due to the real-time response of microblog, individuals like to use microblog sharing the topics, which happened around them, especially in news headlines. As the rapid growth of users and topic numbers, tracking the progress of topics has become a must. However, topic drift and an ocean of noise are common seen in microblog stream. In order to address the problems of topic tracking, we propose an algorithm LMT based on language model and inference network, and adopt the microblog entropy to weigh the importance of each microblog. 12 million microblogs posted by more than 170 thousand users are collected as out experiment dataset, and the experiment results show that our algorithm is more efficient and less noisy compared with traditional Dynamic Topic Model. ?, 2015, Journal of Computational Information Systems. All right reserved.