Exploring LSTM Networks Approach on TRIP Dataset

Current large pretrained language models such as BERT, RoBERTa have achieved high end performance for language understanding task. However, when it comes to the reasoning process by providing supporting evidence to prove the prediction, the performance is not high as expected.

This project is to design and implement NLP system to judge whether a given story is plausible and identify sentences causing the conflict on TRIP Dataset (storks et al., 2021) to explore the reasoning process of large pretrained models besides their end performance in Pytorch.