Self-Information Loss Compensation Learning for Machine-Generated Text Detection

The technology of automatic text generation by machine has always been an important task in natural language processing, but the low-quality generated seriously affects user experience due to poor readability and fuzzy effective information. machine-generated detection method based on traditional learning relies a large number artificial features with rules. general classification deep tends orientation topics, logical information between texts sequences is not well utilized. For this problem, we propose end-to-end model which uses self-information compensate for loss modeling process, learn detection. This task. We experiment Chinese question answer dataset collected from biomedical social media, includes human-written text. result shows that our exceeds most baseline models.