AcTED: Automatic Acquisition of Typical Event Duration for Semi-supervised Temporal Commonsense QA
This addresses the need for efficient temporal commonsense reasoning in NLP, though it is incremental as it builds on existing semi-supervised and BERT-based techniques.
The paper tackles the problem of acquiring typical event durations for temporal commonsense QA by proposing a voting-driven semi-supervised approach to generate pseudo-labeled data, achieving performance comparable to BERT-based methods with far fewer examples and a 7% improvement in Exact Match over RoBERTa baselines.
We propose a voting-driven semi-supervised approach to automatically acquire the typical duration of an event and use it as pseudo-labeled data. The human evaluation demonstrates that our pseudo labels exhibit surprisingly high accuracy and balanced coverage. In the temporal commonsense QA task, experimental results show that using only pseudo examples of 400 events, we achieve performance comparable to the existing BERT-based weakly supervised approaches that require a significant amount of training examples. When compared to the RoBERTa baselines, our best approach establishes state-of-the-art performance with a 7% improvement in Exact Match.