Social short text sentiment classification method and device and storage medium

Abstract

The invention discloses a social short text sentiment classification method and device and a storage medium, and the method comprises the steps: obtaining a text published by a user when the sentimenttendency of a social short text is automatically analyzed and mined; then extracting and calculating a text context sentiment feature value and a priori polarity sentiment feature value; using pre-training word vector models such as word2Vec and the like for training text word vector representation; using multi-window convolution operation for obtaining text semantic relations of different granularities; using a multi-head self-attention network for effectively capturing and memorizing text long-distance context semantic information, and avoiding the problem that a recurrent neural network cannot conduct parallel computing; using multi-layer convolution and pooling combination for abstracting text sentiment semantic representation from different levels; and finally combining sentiment feature vector representation of different levels, so that explicit and implicit sentiment semantic features contained in the text can be described more comprehensively and diverse, and the performance of short text sentiment polarity analysis is improved.

Description

technical field [0001] The present invention relates to the field of natural language processing, in particular to a social short text sentiment classification method, device and storage medium. Background technique [0002] Emerging social media represented by Weibo, Twitter, and WeChat have become an important medium for Internet users to obtain news information, social interaction, self-expression, share opinions, disseminate information and social participation, and become an important medium for social public opinion, corporate brand and product promotion. And the main platform of traditional media communication. As of the first quarter of 2020, the monetizable monthly active users of Twitter, a global microblogging service site, reached 120 million, and the monthly active users of Sina Weibo, a Chinese microblogging platform, reached 550 million. Hundreds of millions of active users are generating a large amount of information containing users' subjective opinions and...

AI Technical Summary

Problems solved by technology

[0005] In view of the above-mentioned existing short text classification methods, it is difficult to describe the long-distance context semantics and the calculation speed of the recurrent neural network structure method is slow, and parallel calculations cannot be performed.

Images