CIMAWA : Entwicklung und Anwendung einer textbasierten Assoziations-Berechnungsmethode

The present work discusses the development and application of a novel method for text-based word association measuring. The method is entitled as CIMAWA which stands for the ‘Concept for the Imitation of the Human Ability of Word Association‘. CIMAWA calculates the strength of the relationship between words. Taking into account the human ability of word association as an archetype, CIMAWA is aimed at simulating the existing but not necessarily discovered associations. It applies statistical analysis to detect co-occurring terms and frequencies based on huge collections of texts, and uses the outcomes for the calculation of the strength of the relation.
CIMAWA is verified in several case studies, especially in comparison with free association tests of human test subjects. In addition the literatures of association measuring are reviewed and the most common methods are implemented and compared with CIMAWA’s outcomes. A detailed explanation of the calculation and the parameters are given, as well as a demonstration of the conceptual differences between CIMAWA and other measurement methods.
The multilateral areas of application and the practical adaptability are shown in four independent software applications. The first application shows how CIMAWA is utilized to detect multi-topic structures in text documents. The second discusses the meta-analysis of text documents in maintenance management. The third presents a CIMAWA based recommender system for text documents towards improving quality of industrial goods. Finally, the fourth application is developed for associative search engine in companies.