Comparative study between expert and non-expert biomedical writings: their morphology and semantics.

The amount of health information on the internet is constantly growing but little is done for detecting the technicality level of these documents and guiding of users towards documents which are appropriate to their expertise level. The objective of our work is to propose clues for the automatic distinction between expert and non expert medical documents. More precisely, we propose to study their morphological and semantic levels. We apply NLP tools, which provide access to the morpho-semantic content of documents. The work is done with French documents within three medical fields (cardiology, pneumology, diabetes). Our experiments and results highlight the fact that this level can indeed provide clues for the distinction of the technicality of documents, and that they appear to be significant and stable across the studied medical fields.