Évaluation d'un générateur automatique de reformulations médicales

Download PDF
Open Webpage

Ioana Buhnila, Amalia Todirascu

Published: 20 Jun 2023, Last Modified: 22 Apr 202418e Conférence en Recherche d'Information et Applications\\16e Rencontres Jeunes Chercheurs en RI\\30e Conférence sur le Traitement Automatique des Langues Naturelles\\25e Rencontre des Étudiants Chercheurs en Informatique pour le Traitement Automatique des LanguesEveryoneCC BY-NC 4.0
Abstract: Evaluation of an automatic medical paraphrase generator. Medical texts are difficult to understand for laypeople due to technical terms. Medical concepts need to be paraphrased using words from the common language. Paraphrasing is the process of rewriting with the aim of explaining or simplifying a sentence or phrase. We present the method for manually building a dataset for paraphrase detection and generation. We are conducting experiments on automatic generation of sub-phrastic medical paraphrases with the machine learning tool APT (Nighojkar & Licato, 2021), using deep learning techniques. We train the Transformer T5 language model (Raffel et al., 2020) with manually annotated medical terms and paraphrases in French and Romanian, a Romance language lacking resources for NLP applications. We present a detailed analysis of the results of the paraphrase generation.