Summary of Realmedqa: a Pilot Biomedical Question Answering Dataset Containing Realistic Clinical Questions, by Gregory Kell et al.
RealMedQA: A pilot biomedical question answering dataset containing realistic clinical questions
by Gregory Kell, Angus Roberts, Serge Umansky, Yuti Khare, Najma Ahmed, Nikhil Patel, Chloe Simela, Jack Coumbe, Julian Rozario, Ryan-Rhys Griffiths, Iain J. Marshall
First submitted to arxiv on: 16 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Clinical question answering systems aim to provide timely answers to clinicians’ queries. Despite advancements, adoption in clinical settings has been slow. One hurdle is the lack of datasets reflecting real-world needs of health professionals. We present RealMedQA, a dataset generated by humans and an LLM (Large Language Model). Our process ensures high-quality QA pairs. We evaluate various QA models on BioASQ and RealMedQA to assess their performance in matching answers to questions. Results show that the LLM is more cost-effective for generating ideal QA pairs. Additionally, our dataset’s lower lexical similarity between questions and answers poses a challenge to top-performing QA models, as evident from the results. We publicly release our code and dataset to facilitate further research. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Clinical question answering systems are meant to provide doctors with helpful answers quickly. Despite progress, these systems haven’t been widely adopted in hospitals yet. One problem is that there aren’t enough datasets that show what kind of questions healthcare professionals really need help with. Our team created a new dataset called RealMedQA that has realistic clinical questions generated by humans and a computer program. We tested different ways for machines to answer questions on this dataset and BioASQ, another popular question-answering benchmark. The results showed that the computer program is more efficient at creating perfect question-answer pairs. Our dataset also makes it harder for top-performing question-answering systems to find answers because the questions are less similar to each other. |
Keywords
» Artificial intelligence » Large language model » Question answering
