Method, device, electronic device and readable storage medium for document sentiment analysis

Abstract

Embodiments of the present invention provide a document sentiment analysis method, device, electronic equipment and readable storage medium, which can help improve the analysis effect and make sentiment analysis closer to daily life. The method includes: obtaining the document, and preprocessing the document to obtain the clauses and words of the document; establishing the index relationship between the clause and the document, and the word and the document; using the topic sentiment model to model the clause and the word, and generating the document The emotion of each clause and the theme of each word, and establish the correspondence between clauses and words; according to the emotion of each clause, the theme of each word, the correspondence between clauses and words, and clauses and documents , the index relationship between words and documents, and calculate the probability distribution of "document-emotion-clause" and the probability distribution of "document-subject-word"; according to the probability distribution of "document-emotion-clause" and the " document-topic-word" to calculate the sentiment orientation of a document.

Description

technical field [0001] The present invention relates to the technical field of the Internet, in particular to a method, device, electronic equipment and readable storage medium for document sentiment analysis. Background technique [0002] With the rapid development of Internet technology, various Internet applications such as new media and e-commerce have been popularized, such as social networks, Weibo, WeChat, e-commerce, etc. Information in these fields is increasingly affecting all aspects of people's lives. Taking the field of e-commerce as an example, consumers' shopping habits have quietly shifted from offline to online. The popularity of online shopping has brought about an explosion of product review information, which has become an indispensable part of e-commerce. component. By analyzing the dissemination information of social networks, comments on Weibo and WeChat, and document information such as commodity review information in the e-commerce field, it is easi...

AI Technical Summary

Problems solved by technology

[0006] (1) The determination of emotional feature words depends on specific field information, and the adaptability of the algorithm is poor

[0007] (2) The granularity of sentiment analysis is large, and the basic object of analysis is the whole sentence or even the whole document. In real life, a document may have multiple sentiment tendencies

[0008] (3) It is necessary to manually label the information, which will consume a lot of labor costs

Moreover, due to the different cognitive levels of different annotators, the annotation results may also be different

[0009] (4) Lack of feedback mechanism, the existing technology lacks the re-learning of wrong judgment results, or the process of re-enriching the unsampled lexicon

Images