Keywords Extraction Algorithm of Financial Review Based on Dirichlet Multinomial Model

Abstract

At present, the main method of keyword extraction is usually based on topical discovery model, and the most popular model is LDA (Latent Dirichlet Allocation) topical model. However, LDA model has poor performance in dealing with short texts, which is the imperfection of the current topical discovery model. This paper propose a keywords extraction model SDMM (Statistical Dirichlet Multinomial model) to extract the keywords of short financial review. This keywords extraction model use DMM (Dirichlet Multinomial Model) for topic discovery, and calculate the similarity between the words and keywords, thus improve the performance of keyword extraction for short texts. Experimental results on several financial review datasets show that the proposed model is better than the existing models in evaluation indicator Precision, Recall and F-measure.