Abstract:

Latent Dirichlet Allocation(LDA) model is one of the important methods in topic mining.To understand the research progress in this field,the paper comprehensively expounds and analyzes the research results of this model in the five years from 2012 to 2016.The paper introduces the LDA model and summarizes the extended model research in seven aspects: lexical, topic evolution, hierarchical relationship, emotion analysis, short text, label and comparative text mining.The applications of this model in topic exploration, recommendation system, forecasting system, filtering system and image processing are summarized.Finally,the future research in this field is forecasted, including improving the evaluation system, improving the precision and efficiency of the model, and introducing the visualization technology into the model.

Key words: Latent Dirichlet Allocation(LDA), Probability topic model, Topic modeling