Aspect-level sentiment analysis method, system and model based on double attention mechanism

Abstract

The embodiment of the invention provides an aspect-level sentiment analysis method, system and model based on a double attention mechanism, and belongs to the technical field of natural language processing. The method comprises the steps of obtaining a to-be-recognized text; mapping the text into continuous word vectors of a low-dimensional space; obtaining a forward hidden vector and a backward hidden vector; splicing the forward hidden vector and the backward hidden vector to obtain a hidden vector; obtaining an external attention weight of the hidden vector; obtaining an internal attention weight of the hidden vector; obtaining a comprehensive attention weight according to the external attention weight and the internal attention weight; determining a first vector; selecting a vector with a relatively large comprehensive attention weight from the first vectors to obtain a second vector; performing convolution operation on the second vector to obtain a third vector; selecting the vector with the maximum value from the third vectors to obtain a fourth vector; adopting a softmax function to calculate the probability that the text is classified into each category; and selecting the category with the maximum probability as a classification result of the text.

Description

technical field [0001] The present invention relates to the technical field of natural language processing, in particular to an aspect-level sentiment analysis method, system and model based on a dual attention mechanism. Background technique [0002] Sentiment analysis is a subfield of natural language processing with a wide range of real-life application scenarios. For example, companies can conduct sentiment analysis on Taobao product evaluations, Ele.me takeaway evaluations, etc., to determine the emotional tendencies contained in them, so as to better understand user needs and promote product updates and iterations. Aspect-level sentiment analysis is a fine-grained sentiment analysis task that aims to analyze the sentiment polarity of text for different aspects. For example, in the comment "Staffs are not that friendly, but the taste covers all", for the aspect word "service", the sentence shows a negative emotional polarity. And for the aspect word "food", the senten...

AI Technical Summary

Problems solved by technology

[0006] First, the existing methods generally use glove or word2vec to encode word vectors for sentences and aspect words in terms of feature representation. The word vectors trained in this way are all static word vectors, which cannot solve the problem of polysemy.

[0007] Second, the traditional method considers single-level attention information, that is, only considers the relationship between words within a sentence or the relationship between aspect words and sentences, and cannot dig deep attention information.

[0008] Third, the CNN-type model cannot obtain context information, while the RNN-type sequence model has the characteristics that the training time is too long and the dependence between words gradually weakens as the distance increases.

Previous methods did not combine the advantages of the two well

Images