Aspect-level sentiment word recognition method based on dependency relationship attention model

Abstract

The invention relates to an aspect-level sentiment word recognition method based on a dependency relationship attention model, and belongs to the technical field of sentiment analysis in natural language processing. According to the method, a multi-head attention model is established according to a syntactic dependency relationship, and the multi-head attention model is used for aspect-level sentiment word recognition. First, a syntactic dependency structure is generated using a syntactic analysis tool. Then, considering that the syntactic dependency structure generated by the tool has a certain error, correcting the syntactic dependency structure; and finally, constructing a multi-head attention model based on a syntactic dependency relationship, and using the multi-head attention model for aspect-level sentiment word recognition. Compared with the prior art, the emotion word recognition precision ratio, the recall rate and the F1 value are remarkably improved under the same environment and the same data set.

Description

technical field [0001] The invention relates to an aspect-level emotional word recognition method based on a dependency relationship attention model, which belongs to the technical field of emotional analysis in natural language processing. Background technique [0002] Aspect-level sentiment word recognition refers to fine-grained sentiment word recognition for user comments on review websites. Different from the traditional coarse-grained emotional word recognition (recognizing emotional words in a sentence) technology, the aspect-product emotional word recognition performs emotional word recognition for specific aspect words. For example, for the sentence "This restaurant tastes average, but the service is attentive", for "taste" (aspect word), identify "average" (emotional word), and for "service", identify "attentive" . This technology is of great significance for merchants to improve their products, or for users to know information about products. [0003] For aspec...

AI Technical Summary

Problems solved by technology

In the GCN-based method, although the syntactic dependency structure is introduced, it does not take into account the impact of different dependency types on emotional word recognition, and the type of syntactic dependency is crucial to the aspect-level emotional word recognition task.

Images