Semantic Coherence, Syntactic Structure, and Relevance Scale (SCSRS): A Tool for Evaluating Language Generation Models

Abstract: This study introduces the Semantic Coherence, Syntactic Structure, and Relevance Scale (SCSRS) as a method for evaluating language generation models. Traditional metrics often fail to fully capture the nuances of language generation, and the SCSRS provides a structured, multi-dimensional approach for assessing model performance. By evaluating semantic coherence, syntactic structure, and relevance to prompts, the SCSRS offers a robust tool for assessing the quality and usability of generated responses. This paper outlines the development of the SCSRS, including its conceptual foundation, example generation, pilot testing, and scoring methodology.

Keywords: Verbal Behaviour. Artificial Intelligence. Prompt Engineering.