Span-based fine-grained sentiment analysis method

Abstract

The invention provides a span-based fine-grained sentiment analysis method, and relates to the technical field of sentiment analysis. According to the invention, a sentiment analysis model based on aspects is established by selecting a sentiment analysis data set, the sentiment analysis model based on aspects is trained through a loss function and a training data set, and sentiment analysis of thetext is achieved by making the test data set to be subjected to sentiment analysis pass through the trained sentiment analysis model based on aspects. The network model provided by the invention is used for extracting aspects and corresponding sentiment polarities, and the model is a simple and effective joint model for extracting sentence aspects and corresponding sentiment polarities for sentiment analysis tasks. The model uses BERT as word embedding, then uses a loop control unit to extract the representation of each subtask, uses an interaction layer to consider the relationship between them, and finally performs aspect item extraction and sentiment analysis.

Description

technical field [0001] The invention relates to the technical field of sentiment analysis, in particular to a span-based fine-grained sentiment analysis method. Background technique [0002] With the development of the Internet age, social networking based on network platforms has become an indispensable part of people's lives. Users are no longer satisfied with obtaining information unilaterally, but actively create information. More and more users are keen to express their emotions, opinions and attitudes on the Internet, such as sharing their viewing experience of a certain movie on a movie website, expressing their opinions on something on a social network, and sharing their views on a certain thing on a shopping website. Post a review on a product, etc. For a large amount of comment data, it is difficult for people to quickly extract information that is effective for them. Therefore, sentiment analysis (Sentiment analysis) came into being. Sentiment analysis, also k...

AI Technical Summary

Problems solved by technology

When the aspect is composed of multiple words, the existing methods predict the labels of the words separately, which cannot make use of the global information, resulting in incorrect extraction of aspect items

And predicting the word labels separately may lead to different emotional information for a certain aspect learned by different words, and the emotional consistency of the aspects cannot be guaranteed.

Images