Aspect-level text sentiment analysis method based on heterogeneous graph neural network

Abstract

The invention discloses an aspect-level text sentiment analysis method based on a heterogeneous graph neural network, and belongs to the field of language processing. The method comprises the following steps: constructing a word-sentence-evaluation aspect three-level graph structure network according to a co-occurrence relationship between words and sentences in a text and evaluation aspects contained in the sentences; then obtaining an initial embedded vector representation of each node; training model parameters by using a graph attention network, continuously updating embedded vector representation of nodes in the graph network according to a connection relation of the nodes in the graph network through a multi-head attention mechanism, and finally predicting aspect-level emotional tendency of the text; and according to the finally obtained embedding vector representation of the sentence node and the evaluation aspect node, calculating the correlation between the sentence node and the evaluation aspect node by using a self-attention mechanism, thereby obtaining the predicted text aspect level emotional tendency. According to the method, the expression ability and generalization ability of the model are effectively improved.

Description

Technical field [0001] The present invention belongs to the field of language processing, in particular, as a heterogeneous diagram neural network, a level text emotional analysis method. Background technique [0002] Aspect-based Sentiment Analysis, ABSA is a fine-grained textual emotion analysis method, which provides more detachable emotional information compared to traditional emotional analysis. Aspective emotional analysis is mainly used to analyze the emotional tendency of text in different aspects (active, neutral, negative), which can be divided into two types of sub-tasks: based on aspects - based emotional classification (ATSA) and category-based emotional classification (ACSA). For example, for text "The food is delicious, that is, the waiter's attitude is too bad", ACSA needs to give a positive emotion for the evaluation, "service" to the evaluation "service" gives an exclusive emotion. [0003] Most of the models at this stage are based on the attention mechanism an...

AI Technical Summary

Problems solved by technology

The above four types of technical solutions have the following shortcomings: First, the existing models all assume that the input texts are independent and identically distributed, but the main research object of aspect-level sentiment analysis-commentary texts often have strong correlations , ignoring this correlation will lead to the loss of a large amount of information, making the model performance decline

Second, the existing models ignore the structural similarity between texts with the same emotion for the same evaluation, which leads to the inability to share information between texts, making the expression ability of the model worse

Third, the existing models ignore the diversity of semantic expressions between texts with the same sentiment in the same evaluation aspect. The loss of this diversity information will lead to poor generalization ability of the model.

Images