Towards Understanding of Medical Randomized Controlled Trials by\n Conclusion Generation

Abstract

Randomized controlled trials (RCTs) represent the paramount evidence of\nclinical medicine. Using machines to interpret the massive amount of RCTs has\nthe potential of aiding clinical decision-making. We propose a RCT conclusion\ngeneration task from the PubMed 200k RCT sentence classification dataset to\nexamine the effectiveness of sequence-to-sequence models on understanding RCTs.\nWe first build a pointer-generator baseline model for conclusion generation.\nThen we fine-tune the state-of-the-art GPT-2 language model, which is\npre-trained with general domain data, for this new medical domain task. Both\nautomatic and human evaluation show that our GPT-2 fine-tuned models achieve\nimproved quality and correctness in the generated conclusions compared to the\nbaseline pointer-generator model. Further inspection points out the limitations\nof this current approach and future directions to explore.\n

Ähnliche Arbeiten