Aspect-level sentiment analysis method and device, electronic equipment and storage medium

Abstract

The application relates to an aspect-level sentiment analysis method, which comprises the following steps: obtaining a target sentence to be analyzed and carrying out embedding to obtain a target embedding vector; constructing a template sentence, carrying out embedding operation on template words, aspect words and mask words of the template sentence to obtain template word embedding vectors, aspect word embedding vectors and mask embedding vectors; inputting the template word embedding vectors into a Prompt encoder to obtain template word tensors, splicing the template word tensors, aspect word embedding vectors and mask embedding vectors to obtain continuous template vectors; jointly inputting the target embedding vector and the continuous template vector into a pre-trained language model and carrying out a pooling operation to obtain context features; generating a syntax dependency tree according to the target sentence, extracting syntax dependency features according to the syntax dependency tree; carrying out feature fusion on the context features and the syntax dependency features to obtain fusion features; and predicting according to the fusion features to obtain the sentiment polarity of the aspect words in the target sentence. The accuracy of aspect-level sentiment analysis can be improved.

Images