Self-Adaptive Semantics Masking for Text Classification

Abstract

Text classification is a typical problem in NLP, which is to classify a given text into a certain category. In this article, we propose a classification-based training method, which is mainly based on the input text adaptive masking, and then further pre-training and fine-tuning training. This training method makes it easier to obtain the semantic features of the text, and thus can improve the semantic learning ability of the model, and thus can improve the classification effect. Experiments show that compared with other classic methods, this method can now improve the classification performance.