Section: Software

IGNG-Fv2

Participant : Jean-Charles Lamirel [correspondent] .

The IGNG-Fv2 program implements a new incremental clustering algorithm whose main domain of application is the statistical analysis of continuous flow of evolving textual data, as well as the one of static textual data. It has been developed in the context of the CPER TALC (McFiiD action). It is based on a generic adaptation of the classical neural-based clustering approaches relying on gas of neurons with free topology. The IGNG-Fv2 approach exploits a combination of distance based and cluster data feature maximization criteria. This approach has been proved more efficient than the usual techniques for the analysis all kinds of static textual datasets. Considering its incremental character, it can also provide the information analysts with precise online detection of topic changes in the curse of a textual information flow.