A Mining Method for Opinion Summarization in Text Data Based on Topic Diversity

Abstract

The present invention provides a text data viewpoint summary mining method based on topic diversity, which includes the following steps: step S1: preprocessing the topic text; step S2: input topic corpus and background corpus; step S3: extract topic corpus set topic attributes; step S4: add emotional polarity to the obtained topic attributes for sentence vectorization; step S5: use the obtained topic attributes as evaluation objects, and use the multi-evaluation object-oriented dynamic word sequence sentiment analysis method to analyze The emotional polarity of the evaluation object contained in the sentence is obtained to obtain the emotional attribute features contained in the sentence, and the feature vectorization of a sentence is carried out; Step S6: use the text sentence feature vector obtained in step S5 to construct a diversity objective function. It can efficiently and accurately obtain opinion summaries of topic texts, and can be applied to larger-scale data set application scenarios.

Description

technical field [0001] The present invention relates to the fields of text summarization and sentiment analysis, and more specifically, relates to a method for generating short opinion summaries with rich user emotional information for massive topic text data of Chinese microblog corpus, and the opinion summaries can accurately cover the text discussed It can be applied to practical application scenarios such as news summaries and commodity review summaries. Background technique [0002] Currently, there are many techniques and methods available for research in the field of opinion summarization. Traditional view summarization models include graph models and ranking models. The representative methods of graph models include Textrank, PageRank, LexRank and other methods. They use sentences as nodes, and a certain relationship between sentences as the weight of edges, and iteratively update and calculate the scores of sentences through the random walk model, so as to realize ...

AI Technical Summary

Problems solved by technology

The disadvantage is that most of the existing models consider using the diversity of all words in the text sentence to ensure that the opinion summary covers the main idea of ​​the text, and the diversity of the words is used to ensure the diversity of the summary, but the diversity of words does not guarantee that the opinion summary includes The subject of the source text, words that are not related to the subject will affect the final opinion summary, and the existing research methods describe the emotional information of the summary through the emotional information of the entire text sentence, and the emotions of many irrelevant text subjects are also taken into account. The two factors lead to the final summary containing a lot of content and emotional information that are not related to the main theme of the text

Images