Micro-blog topic drift detection based on VSM and LDA models
HU Xiu-l · Journal of Lanzhou University of Technology · 2015
Aimed at the feature of micro-blog topic that it is ease to drift,its detection is conducted with vector space model,combined into LDA model.Principally,the Gibbs sampling algorithm is used to obtain the probability distribution of every micro-blog words and measure their correlation,identify topic boundary with dynamic constant method,extract topic words by means of computation of lexical information entropy in topic field,and,finally,a topic vector space model is generated.By comparing the word sequence in the topics within the topic vector space model to that within discrete time sequence model,the topic drift detection is realized.It is shown by test that the VSM/LDA model-based micro-blog topic drift detection is a set of effective method.