Med-EASi: Finely Annotated Dataset and Models for Controllable Simplification of Medical Texts

Abstract

Automatic medical text simplification can assist providers with patient-friendly communication and make medical texts more accessible, thereby improving health literacy. But curating a quality corpus for this task requires the supervision of medical experts. In this work, we present Med-EASi (Medical dataset for Elaborative and Abstractive Simplification), a uniquely crowdsourced and finely annotated dataset for supervised simplification of short medical texts. Its expert-layman-AI collaborative annotations facilitate controllability over text simplification by marking four kinds of textual transformations: elaboration, replacement, deletion, and insertion. To learn medical text simplification, we fine-tune T5-large with four different styles of input-output combinations, leading to two control-free and two controllable versions of the model. We add two types of controllability into text simplification, by using a multi-angle training approach: position-aware, which uses in-place annotated inputs and outputs, and position-agnostic, where the model only knows the contents to be edited, but not their positions. Our results show that our fine-grained annotations improve learning compared to the unannotated baseline. Furthermore, our position-aware control enhances the model's ability to generate better simplification than the position-agnostic version. The data and code are available at https://github.com/Chandrayee/CTRL-SIMP.

A text style transfer system for reducing the physician–patient expertise gap: An analysis with automatic and human evaluations

“…For these reasons, we evaluated our outputs and those of the models presented by Cao et al (2020) using several other metrics, referred to both input and target sentences. Computing these metrics for the gold source and target texts allowed us to highlight the degree of content changes in the test set, as suggested in previous works (Basu, Vasu, Yasunaga, Kim, & Yang, 2021;Cao et al, 2020;Vásquez-Rodríguez, Shardlow, Przybyła, & Ananiadou, 2021), and confirmed through our human evaluations. This issue may stem from the loss of contextual information when working at a sentence level.…”

An exploration of knowledge‐organizing technologies to advance transdisciplinary back pain research

“…New LLMs since 2018 (e.g., chatGPT, GPT 4, T5, BART) can generate a wide variety of text analyses and dialogues with an impressive level of fluency out‐of‐the‐box. Through fine‐tuning, LLMs become specialized at particular tasks, such as analyzing social determinants of health from clinical notes, 15 answering disease‐specific questions based on medical literature, 16,17 simplifying medical concepts and texts for patients, 18 and more 14 . Using end‐user‐facing LLM interfaces (e.g., Open AI Playground) with and without AI technical training can improve LLM outputs by prepending prompts—textual instructions and examples of their desired interactions—to LLM inputs.…”

“…Similarly, building a system that explains LBP‐related concepts to out‐of‐domain scientists or stakeholders can take little to no training data. One might only need to adapt such explanation systems from other medical domains with sets of LBP examples 18 …”