NapSS: Paragraph-level Medical Text Simplification via Narrative Prompting and Sentence-matching Summarization

Findings of the Association for Computational Linguistics: EACL 2023, pages 1079–1091 May 2-6, 2023 ©2023 Association for Computational Linguistics NapSS: Paragraph-level Medical Text Simplification via Narrative Prompting and Sentence-matching Summarization Junru Lu1, Jiazheng Li2, Byron C. Wallace3, Yulan He1,2,4 and Gabriele Pergola1 1Department of Computer Science, University of Warwick, UK 2Department of Informatics, King’s College London, UK 3Northeastern University, USA 4The Alan Turing Institute, UK {junru.lu, gabriele.pergola}@warwick.ac.uk b.wallace@northeastern.edu, {jiazheng.li, yulan.he}@kcl.ac.uk Abstract Accessing medical literature is difficult for laypeople as the content is written for special- ists and contains medical jargon. Automated text simplification methods offer a potential means to address this issue. In this work, we propose a summarize-then-simplify two-stage strategy, which we call NapSS, identifying the relevant content to simplify while ensuring that the original narrative flow is preserved. In this approach, we first generate reference summaries via sentence matching between the original and the simplified abstracts. These summaries are then used to train an extractive summarizer, learning the most relevant con- tent to be simplified. Then, to ensure the nar- rative consistency of the simplified text, we synthesize auxiliary narrative prompts com- bining key phrases derived from the syntac- tical analyses of the original text. Our model achieves results significantly better than the seq2seq baseline on an English medical cor- pus, yielding 3%∼4% absolute improvements in terms of lexical similarity, and providing a further 1.1% improvement of SARI score when combined with the baseline. We also highlight shortcomings of existing evaluation methods, and introduce new metrics that take into account both lexical and high-level seman- tic similarity. A human evaluation conducted on a random sample of the test set further es- tablishes the effectiveness of the proposed ap- proach. Codes and models are released here: https://github.com/LuJunru/NapSS. 1 Introduction The medical literature is vast, and continues to ex- pand quickly. Most patients (laypeople), however, are unable to access this information because it is written for specialists and so dense and laden with jargon. As the recent ‘infodemic’ has shown, ac- cess to reliable and comprehensible information about citizens’ health is a fundamental need: for example, a European Health Literacy Survey (HLS- EU) reports that "at least 1 in 10 (12%) respondents S1. Two good quality randomized controlled trials involving 754 women were identified. S2. Rapid negative pressure application reduced the duration of the procedure without any evidence of differences in outcomes for the mother or infant. S3. Rapid method of negative pressure application should be recommended for vacuum extraction assisted vaginal delivery. involving 754 participants. the same success rate of vacuum rocedure of 98.2% by both methods (risk ratio (RR) 1.00, 95% confidence Plain-Language Summary - PLS S1. We included two trials involving 754 participants. S2. One new trial of 660 participants showed the same success rate of vacuum procedure of 98.2% by both methods (risk ratio (RR) 1.00, 95% confidence interval (CI) 0.98 to 1.02). S3. The two included trials showed significant reductions in the time between applying the vacuum cup and delivery, (one trial (74 women): mean …

Content truncated to 500 words