Biomedical text readability after hypernym substitution with fine-tuned large language models
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The advent of patient access to complex medical information online has highlighted the need for simplification of biomedical text to improve patient understanding and engagement in taking ownership of their health. However, comprehension of biomedical text remains a difficult task due to the need for domain-specific expertise. We aimed to study the simplification of biomedical text via large language models (LLMs) commonly used for general natural language processing tasks involve text comprehension, summarization, generation, and prediction of new text from prompts. Specifically, we finetuned three variants of large language models to perform substitutions of complex words and word phrases in biomedical text with a related hypernym. The output of the text substitution process using LLMs was evaluated by comparing the pre- and post-substitution texts using four readability metrics and two measures of sentence complexity. A sample of 1,000 biomedical definitions in the National Library of Medicine's Unified Medical Language System (UMLS) was processed with three LLM approaches, and each showed an improvement in readability and sentence complexity after hypernym substitution. Readability scores were translated from a pre-processed collegiate reading level to a post-processed US high-school level. Comparison between the three LLMs showed that the GPT-J-6b approach had the best improvement in measures of sentence complexity. This study demonstrates the merit of hypernym substitution to improve readability of complex biomedical text for the public and highlights the use case for fine-tuning open-access large language models for biomedical natural language processing.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.002 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it