The cause relation in biopharmaceutical corpora: English and French patterns for knowledge extraction.

Abstract

One of the most important aspects of a terminologist's work is extracting conceptual information about terms from texts. Because this task is so time-consuming, researchers are trying to develop tools which will extract conceptual information semi-automatically. Many of these tools are based on the use of linguistic indicators called knowledge patterns. This thesis aims to identify some knowledge patterns in English and French which indicate the conceptual relation of cause and effect. This relation, though not as widely studied as those of generic to specific or part to whole, is critical in many subject fields including medicine and pharmaceuticals. For this reason, our research focuses on biopharmaceutical texts. Our methodology involved building representative corpora in English and French, and then identifying possible knowledge patterns. The precision of the identified patterns was calculated in order to predict their possible effectiveness for semi-automatic knowledge extraction. We discuss some of the issues observed in the process of identifying these patterns, and those which might affect the subsequent implementation of the patterns for semi-automatic knowledge extraction. A brief interlinguistic comparison of the English and French patterns identified is also included. Our research shows that the subject field of biopharmaceuticals contains many potentially productive knowledge patterns for the cause relation. However, there are also many issues which must be taken into account when identifying patterns and developing knowledge extraction tools.