Self-organizing map for conceptual modelling

Abstract

Textual description of the concept plays an important role in the conceptual model comprehensibility. In this paper, the self-organizing map to test the model comprehensibility is suggested. An explanatory text of the concept is transformed into a numerical vector. Several vector spaces are build using the hyponymy of the concepts from the WordNet dictionary. The received conceptual model vector space is tested for self-organisation properties with self-organizing maps. An experiment with the conceptual model self-organizing map and IBM toolbox for natural language understanding shows that it is useful to use them in the systems that support natural language modality. IBM financial services data model (FSDM) was used for the present research.