This study aims to evaluate the effectiveness of a Meta AI tool using a human-machine co-learning model for practicing and learning Taiwanese and English languages between Taiwan and Japan. Teachers and students from both countries followed a six-step human-machine co-learning model that integrated both the human and machine co-learning process in the Taiwanese and English languages to conduct the study. We assessed the performance of both human intelligence and machine intelligence in Taiwanese language practice using the sentence BERT similarity approach, to measure how similar the source sentences were to the Meta AI tool-generated sentences. Human experts then extracted important features of the source sentences to construct a knowledge model, and we employed the proposed evolutionary computation-based assessment model to analyze the similarity in learning effectiveness between human and machine evaluations. Our findings indicate that the Meta AI tool has a positive impact on language practice and learning, and the human-machine co-learning model will be an effective approach for Taiwanese and English language learning between Taiwan and Japan in the future.