Many scientific publications and materials on artificial intelligence (AI) have been written in English; however, for many AI learners, English is their second language. Therefore, the difficulty (readability) of online self-teaching texts on AI for English-as-a-second-language (ESL) learners is essential for determining the language support ESL learners need to learn AI. However, only a few studies have addressed this issue. Therefore, we identified the difficulty level of English self-teaching texts for ESL AI learners. Because large-scale testing for ESL learners is impractical owing to the financial costs and time involved, we built two distinctive automatic readability assessors: one using sophisticated deep-learning-based natural language processing (NLP) technology, and another using classic NLP based on word frequency and applied linguistics. We conducted our evaluation using AI research papers and university-level online course texts. Interestingly, the distinctive automatic assessors, which were trained on different datasets, showed similar results. Intermediate-level ESL learners could read approximately 10% of online course texts. We also showed that they are significantly easier to read than AI research papers for ESL learners, demonstrating their usefulness in AI learning.
CITATION STYLE
Ehara, Y. (2022). Assessing Readability of Learning Materials on Artificial Intelligence in English for Second Language Learners. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 13356 LNCS, pp. 475–478). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-11647-6_96
Mendeley helps you to discover research relevant for your work.