Pentagon at MEDIQA 2019: Multi-task Learning for Filtering and Re-ranking Answers using Language Inference and Question Entailment
Hemant Pugaliya, Karan Saxena, Shefali Garg, Sheetal Shalini, Prashant Kumar Gupta, Eric Nyberg, Teruko Mitamura · 2019
Parallel deep learning architectures like finetuned BERT and MT-DNN, have quickly become the state of the art, bypassing previous deep and shallow learning methods by a large margin.More recently, pre-trained models from large related datasets have been able to perform well on many downstream tasks by just fine-tuning on domain-specific datasets (similar to transfer learning).However, using powerful models on nontrivial tasks, such as ranking and large document classification, still remains a challenge due to input size limitations 1 of parallel architecture and extremely small datasets (insufficient for fine-tuning).In this work, we introduce an end-to-end system, trained in a multi-task setting, to filter and re-rank answers in medical domain.We use task-specific pre-trained models as deep feature extractors.Our model achieves the highest Spearman's Rho and Mean Reciprocal Rank of 0.338 and 0.9622 respectively, on the ACL-BioNLP workshop MediQA Question Answering shared-task.* * Equal contribution, randomly sorted.Karan and Shefali took ownership of the NLI module while Sheetal and Prashant worked on the RQE module.Hemant researched and implemented the Question-Answering system including baseline and multi-task learning.Sheetal and Hemant worked on scraping data from icliniq.Karan and Prashant helped with integration of NLI and RQE module respectively into the multi-task system.1