Using generative AI tools and LARA to create multimodal language learning resources for L2 Icelandic

Description

We utilize ChatGPT-4 and other generative AI tools in developing illustrated multimodal resources for learning Icelandic as a second language (L2) via reading in the online platform LARA. These are illustrated short stories in Icelandic created using specific prompts including information about learner levels and age groups. We present a methodology for creating engaging stories and a way to evaluate the quality and suitability of text and images generated by AI. Additionally, we assess the perception rate of adult learners reading a story at A1 and A2 level in LARA. We conclude that generating suitable prompts greatly assists with getting the desired output; however, this is restricted to the language one works with. Much human editing is still necessary for Icelandic texts to improve their quality in grammar, vocabulary, and cultural aspects, and their suitability for language teaching. Responses from 46 learners indicate that the short stories are relatively easy to read but lack wittiness and some cultural information. Most learners of A1 story were impartial, and some thought it was likely it could have been written by AI, however, most learners of A2 story thought this was unlikely. Integrating the stories into LARA helped with learning new vocabulary.

Period	17 Aug 2023
Event title	EUROCALL 2023: CALL for All Languages 15-18 August 2023, University of Iceland, Reykjavik
Event type	Conference
Location	Reykjavík, IcelandShow on map
Degree of Recognition	International