A novel approach to measuring creative analogical fluency in Chinese using advanced language models

This study introduces a novel approach to assessing creative analogical fluency in the Chinese language context. We developed the Creative Analogical Fluency Vector Offset Method, an algorithm to evaluate fluency in creative analogical reasoning using advanced word embedding models. A Chinese dataset of analogical reasoning questions was constructed and categorized into semantic and syntactic domains. The study involved 150 Chinese undergraduate students who completed a 60-item word analogical reasoning test. We compared the performance of three word embedding models (Word2Vec, BERT, and GPT) in assessing creative analogical fluency. Results demonstrated high accuracy across all models. Notably, our method showed comparable effectiveness in evaluating semantic and syntactic analogical reasoning questions, challenging the assumption of significant differences between these domains in the Chinese context. This research contributes to the field by providing a more efficient and culturally relevant tool for assessing creative analogical reasoning.