aiSee: Term Variants Graph

Term variants graph

A term classes/term variants graph as produced by TermWatch, a data mining system. Courtesy of F. Ibekwe-SanJuan, Université Jean Moulin Lyon 3, and E. SanJuan, Univer­sité de Metz.

“My research interests revolve around automated systems for processing textual data for various applications, such as scientific and technological watch, termi­nol­ogy update and structuring, intelligent information retrieval and text mining.

For these applications, I integrate shallow linguistic knowledge into a clustering algorithm which produces classes based on syntactic relations (head/modifier variations). The lin­guistic units used for clustering are term candidates extracted from raw texts. The vari­a­tion search programs and the single-link clustering algorithm are written in Awk. They first build graphs of variants which are later clustered into classes using alternatively variation relations dealing with modifier and head words in a term.

These classes are then visualized using aiSee. They ideally depict the current research topics in a field. Particular attention is paid to needs for thematic trends” survey for sci­en­tific and technological watch. The powerful aiSee interface allows for an intelligent exploration of the internal and external structures of classes.

Fidelia Ibekwe-SanJuan, Université Jean Moulin Lyon 3

White papers