Unsupervised Method for the Acquisition of General Language Paraphrases for Medical Compounds
Natalia Grabar, Thierry Hamon · 2014
Medical information is widespread in modern society (e.g.scientific research, medical blogs, clinical documents, TV and radio broadcast, novels).Moreover, everybody's life may be concerned with medical problems.However, the medical field conveys very specific and often opaque notions (e.g., myocardial infarction, cholecystectomy, abdominal strangulated hernia, galactose urine), that are difficult to understand by lay people.We propose an automatic method based on the morphological analysis of terms and on text mining for finding the paraphrases of technical terms.Analysis of the results and their evaluation indicate that we can find correct paraphrases for 343 terms.Depending on the semantics of the terms, error rate of the extractions ranges between 0 and 59%.This kind of resources is useful for several Natural Language Processing applications (i.e., information extraction, text simplification, question and answering).