A Genetic Algorithm Based Approach for Data Fusion at Grammar Level

Multimodal interaction is a type of Human-Computer Interaction, which involves a combination of multifarious modalities to effectuate a task. Human-to-human interactions necessitate the use of all modalities such as speech, gestures, facial expressions and assorted media available for brainy communications. The single modalities often introduce the ambivalent interpretation of the ideas being conveyed. Multimodal interaction plays a major role in resolving such ambiguities. The various modalities used can be combined as a way to recover the semantics of the messages involved. This strategy, known as data fusion is the main focus of the research presented. This paper proposes a new genetic algorithm based approach for achieving intelligent data fusion for inferred contextfree grammars. Grammar inference methods are applied to generate the grammars, and the resulting production rules are fused to generate a correct grammar. Related results, their implications, and future work are presented.