Mapping the structure of research topics through term variant clustering: the TermWatch system

A multi-disciplinary approach integrating computational linguistic techniques is necessary to elaborate indicators of research topic evolution. We describe a system which bases clustering on linguistic relations, instead of the usual co-occurrence paradigm. The interesting features of this approach, embodied in the TermWatch system, lie in the combination of state-of-the-art techniques in computational terminology, mathematics (graph formalism) and visualization techniques. Computational terminology enable us to extract meaningful text chunks and to relate these chunks through linguistic relations. These text chunks are terms and the linguistic relations are syntactic variations. We integrated into this system an adapted visualization tool which enhances comprehension of the research topic layout and their trends. Here we focus on the chronological analysis of graphs issued by TermWatch through a graph visualization tool, Aisee which helps the end-user to track the main tendencies of research topics in his/her field.