TransBERT: A Three-Stage Pre-training Technology for Story-Ending Prediction

Recent advances, such as GPT, BERT, and RoBERTa, have shown success in incorporating a pre-trained transformer language model and fine-tuning operations to improve downstream NLP systems. However, this framework still has some fundamental problems in effectively incorporating supervised knowledge from other related tasks. In this study, we investigate a transferable BERT (TransBERT) training framework, which can transfer not only general language knowledge from large-scale unlabeled data but also specific kinds of knowledge from various semantically related supervised tasks, for a target task. Particularly, we propose utilizing three kinds of transfer tasks, including natural language inference, sentiment classification, and next action prediction, to further train BERT based on a pre-trained model. This enables the model to get a better initialization for the target task. We take story-ending prediction as the target task to conduct experiments. The final results of 96.0% and 95.0% accuracy on two versions of Story Cloze Test datasets dramatically outperform previous state-of-the-art baseline methods. Several comparative experiments give some helpful suggestions on how to select transfer tasks to improve BERT. Furthermore, experiments on six English and three Chinese datasets show that TransBERT generalizes well to other tasks, languages, and pre-trained models.