Emotion analysis and generation method for news comments

Abstract

The invention discloses an emotion analysis and generation method for news comments, and the method specifically comprises the following steps: preparing an initial data set which comprises a manually labeled small-scale data set A and a large-scale unsupervised data set B; on the basis of the data set B, using an open source sentiment classification tool to establish a large-scale pseudo-tag data set C for the news comment sentiment analysis model; pre-training a news comment sentiment analysis model based on the large-scale pseudo-tag data set C; after pre-training, performing fine tuning on the manually labeled small-scale data set A to obtain a final news comment sentiment analysis model; using the final news comment sentiment analysis model to carry out pseudo label labeling on the data set B again to obtain a large-scale pseudo label data set D; pre-training a controllable news comment generation model based on the large-scale pseudo-tag data set D; and after pre-training, carrying out fine tuning on the manually labeled small-scale data set A to obtain a final controllable news comment generation model.

Description

technical field [0001] The invention belongs to the field of sentiment analysis and text generation in natural language processing, and in particular relates to a method for sentiment analysis and generation of news comments. Background technique [0002] News comments are one of the important windows to listen to the voices of the people and understand their appeals. Faced with the massive news comments generated by the Internet all the time, natural language understanding technology is used to automatically analyze and monitor the mass news comments in real time. Public sentiment and detection of malicious and false comments are of great significance to policy formulation and risk control. While monitoring public opinion, use natural language generation technology to automatically and controllably generate comments for news, which is beneficial to guide the correct direction of public opinion and control negative public opinion. [0003] Sentiment analysis can be abstract...

AI Technical Summary

Problems solved by technology

[0005] However, the existing technology still has the following problems: when the existing pre-trained language model is migrated to the two downstream tasks of news comment sentiment analysis and generation, in order to ensure the effect, it still relies on a large amount of labeled data, and there is still a lack of high-level training in the field. High-quality relevant data sets, and manual labeling is time-consuming and laborious; most of the existing news comment sentiment analysis methods treat comments in isolation, without considering the semantics of the original news text at the same time; most of the existing comment generation methods are weak in controllability, only based on The original news text randomly generates comments, and it is impossible to set the desired emotional polarity at the time of generation to achieve public opinion guidance and other purposes; and research usually separates the two tasks of news comment sentiment analysis and generation, and lacks mutual promotion of the two tasks. relationship considerations

Images