Accuracy of Natural Language Processing Algorithms in Representing Stressed Words in Closed Captioning

Abstract

For English as a Foreign Language (EFL) learners whose native tongue is not a stress-timed language, it is a difficult task to master word stress in speaking. Despite technological advances and years of learning English in school, many Thai EFL learners, in particular, are struggling with correct stress placement in English words, especially when conversing in English. With knowledge of natural language processing and English linguistics, this study creates closed captions to provide additional visual information on stress placement to help these learners making themselves better understood in English conversation. This also helps those who are deaf or hard of hearing to recognize highlighted audio narratives properly. The purpose of this study is to determine the accuracy of word stress representation in closed captions via human auditory analysis and perception. The results show that the current approach delivers 74.19% of representation accuracy.