Neural network, training method, aspect-level sentiment analysis method and device and storage medium

Abstract

The invention discloses a neural network, a training method, an aspect-level sentiment analysis method and device and a storage medium. The neural network comprises the steps that a BERT word embedding module obtains semantic information of an input text and word vectors of all words of a to-be-analyzed statement in the input text; the first full connection layer obtains a semantic feature vector based on the semantic information; the second full connection layer obtains word feature vectors based on the word vectors; a potential Dirichlet distribution module extracts topic distribution of a statement to be analyzed and topic distribution of an aspect emotion pair; the feature fusion layer judges whether the to-be-analyzed statement is matched with the aspect emotion pair or not; a decoding marking module marks the position of the target word in the decoding sequence to obtain a marking sequence; and the output layer outputs an emotion analysis result based on the judgment result output by the feature fusion layer. And multi-task output is realized, semantic information and topic distribution are subjected to feature fusion through the feature fusion layer, vocabularies in a specific field are learned, and the sentiment classification task effect of the model is improved.

Description

technical field [0001] The invention relates to the fields of natural language processing and artificial intelligence, in particular to a neural network, a training method, an aspect-level sentiment analysis method, a device and a storage medium. Background technique [0002] With the vigorous development of Internet technology, no matter whether people are shopping, socializing or choosing catering services on the Internet, they are all keen to express their opinions, resulting in a large number of online comments. And these text-based comments contain people's emotional expressions on a certain entity and different aspects of the entity. By performing sentiment analysis on user comments, companies can clarify the defects and deficiencies of their products or services, so that they can improve their products or services in a more targeted manner to enhance the competitiveness of the company, and users can also get more in line with The product or service of your own mind w...

AI Technical Summary

Problems solved by technology

[0007] To sum up, most of the currently published patents and literature on aspect-level sentiment analysis do not consider how to better construct word vectors in specific domains, and also consider the difficult samples of the model and the learning of low-frequency words. Therefore, an aspect-level sentiment analysis method to solve such problems is urgently needed in this field

Images