TLDR at SemEval-2024 Task 2: T5-generated clinical-Language summaries for DeBERTa Report Analysis
This work addresses the problem of improving entailment and contradiction analysis in clinical trials for researchers, though it is incremental as it builds on existing models like T5 and DeBERTa.
The paper tackled the challenge of small context windows and lengthy premises in clinical natural language inference by generating T5-based premise summaries for DeBERTa analysis, resulting in a 0.184 increase in Macro F1 scores over truncated premises.
This paper introduces novel methodologies for the Natural Language Inference for Clinical Trials (NLI4CT) task. We present TLDR (T5-generated clinical-Language summaries for DeBERTa Report Analysis) which incorporates T5-model generated premise summaries for improved entailment and contradiction analysis in clinical NLI tasks. This approach overcomes the challenges posed by small context windows and lengthy premises, leading to a substantial improvement in Macro F1 scores: a 0.184 increase over truncated premises. Our comprehensive experimental evaluation, including detailed error analysis and ablations, confirms the superiority of TLDR in achieving consistency and faithfulness in predictions against semantically altered inputs.