Emotion prediction model training method and device

Abstract

The invention discloses an emotion prediction model training method and device. The training method comprises the following steps: acquiring a plurality of unsupervised corpora; and covering at least one emotion expression in the unsupervised corpus, and inputting the corpus before covering and the corpus after covering into an emotion prediction model as training samples. The emotion prediction model is used for performing a word prediction task and a word emotion prediction task for the masked emotion expression in the training sample, and calculating a task error according to a word prediction task result, a word emoition prediction task result and the corpus before masking, and correcting a weight coefficient of the emotion prediction model based on the task error. According to the method, the emotion expression is masked to construct the training samples, and the to-be-trained emotion prediction model is trained by using the training samples, so that the accuracy of an emotion prediction analysis task can be improved, the stability of the model can be improved, and the number of the training samples is greatly reduced.

Description

technical field [0001] The present disclosure relates to the field of natural language processing, in particular, to a training method and device for an emotion prediction model. Background technique [0002] In recent years, pre-trained language models have been widely used in multi-domain sentiment analysis and achieved good results. However, fine-tuning the model in the source domain is prone to overfitting due to the variability of user expressed emotions in different domains, resulting in poor performance in the target domain when transferring the model from the source domain to the target domain. Moreover, due to the large parameters of the pre-trained model, fine-tuning consumes a lot of training speed and space, and requires a higher number of training samples. Contents of the invention [0003] In view of this, the purpose of the present disclosure is to provide a method and device for training an emotion prediction model, so that the weight parameters of the obt...

AI Technical Summary

Problems solved by technology

However, fine-tuning the model in the source domain is prone to overfitting due to the variability of users' expressed emotions in different domains, resulting in poor performance in the target domain when the model is transferred from the source domain to the target domain.

Moreover, due to the large parameters of the pre-training model, fine-tuning consumes a lot of training speed and space, and requires a higher number of training samples.

Images