Efficient Detection of Generative Chinese Statements Using Feature Compression Methods

With the explosion of big models, compliance detection of generative utterances becomes increasingly important. Currently, the detection of Chinese corpus faces the problems of 1. a small number of effective datasets and 2. poor learning effect of Chinese corpus. In this paper, we explore a chinese statement learning approach based on a small amount of data based on the COLD dataset, which proves that the small sample data can also bring a good learning effect to the model under low-dimensional features.