Using Biomedical Terminologies to extract Noun Phrases for managing knowledge evolution

In order to identify variations between two or several versions of Clinical Practice Guidelines, we propose a method based on the detection of noun phrases. Currently, we are developing a comparison approach to extract similar and different elements between medical documents in French in order to identify any significant changes such as new medical terms or concepts, new treatments etc. In this paper, we describe a basic initial step for this comparison approach i.e. detecting noun phrases. This step is based on patterns constructed from six main medical terminologies used in document indexing. The patterns are constructed by using a Tree Tagger. To avoid a great number of generated patterns, the most relevant ones are selected by choosing those that identify more than 80% of the six terminologies used in this study. These steps allowed us to obtain a manageable list of 262 patterns which have been evaluated. Using this list of patterns, 708 maximal noun phrases were found, among them, 364 are correct which represent a 51.41% precision. However by detecting these phrases manually, 602 maximal noun phrases were found which represent a 60.47% recall and by consequence a 55.57% F-measure. We tried to improve these results by increasing a number of patterns from 262 to 493. We obtained a total of 729 maximal noun phrases, with 365 which were correct, which corresponding to a 50.07% precision, 60.63% recall and 54.85% F-measure.