Communication Dans Un Congrès Année : 2025

Creating Multimedia Resources for Kanak Languages Using the C-LARA Platform: Case Studies in Iaai and Drehu

Résumé

We describe an ongoing project where a diverse team, including linguists, ethnomusicologists, applied linguists, language teachers, and computer scientists, is developing multimedia resources for two Kanak languages of New Caledonia—Iaai and Drehu—using the ChatGPT-based Learning And Reading Assistant (C-LARA) platform. C-LARA (https://www.c-lara.org/; (Bédi et al., 2023, 2024)) is an open source platform powered by artificial intelligence (AI) enabling non-expert users to rapidly create illustrated multimodal content for language learners. We present initial proof-of-concept resources already completed and outline plans for further work. For Drehu, the language of the island of Lifou and one of the languages of Tiga, and the Kanak language with the largest community of speakers (approx. 16,000), we developed an interactive alphabet book and a C-LARA edition of Leu me Jö (The (North) Wind and the Sun), a Drehu version of Aesop’s fable. These resources feature Drehu words with accompanying AI-generated images and audio pronunciations recorded by native speakers. The platform provides phonetic annotations using pronunciation respellings based on French orthography (franétique transcriptions, derived from français + phonétique), aiding learners in correct pronunciation and mitigating cross-language orthographic conflicts (see Figure 1). The grapheme-phoneme correspondences were based on the conventions of the Académie des Langues Kanak’s Proposition d’Écriture for Drehu (Sam, 2009) as well as our own research (Anonymous). In Iaai, spoken by around 3,700 speakers mainly on the island of Ouvéa, we created a multimedia version of Wanakat Kaori, a bilingual Iaai-French tale (see Figure 2) adapted from traditional oral literature (Hombouy et al., 2005), as well as several children’s songs, lullabies, and contemporary music pieces. For Wanakat Kaori, we used illustrations from the original storybook, while AI-generated images were created for the other texts. All resources include audio recordings by native speakers. We collaborated closely with the Académie des Langues Kanak (ALK) to ensure adherence to Iaai writing norms (2020) and cultural appropriateness. One challenge was addressing limitations in using AI to create images depicting specific cultural contexts, which we mitigated through prompt engineering and community feedback. Here, the involvement of the ALK and community members was pivotal in refining the resources, ensuring they were culturally relevant and accurately represented the languages. Another challenge was adapting C-LARA to work with Indigenous languages unfamiliar to the AI. Since then, we have enhanced the platform to better handle these challenges by improving the platform’s ability to support manual annotation and glossing in unfamiliar languages and refining the AI-based image generation techniques. Our work demonstrates the potential of AI in supporting linguistic documentation and education for endangered languages. By combining advanced technology with community collaboration, we contribute to the preservation and promotion of Kanak languages, showing how AI tools can be adapted to create culturally relevant multimedia resources for languages with limited digital presence. All materials described are freely accessible online; the final version of this abstract will include links. We are currently developing substantially larger resources for Kanak languages, some of which we expect to have made available by the time of the conference. References Académie des Langues Kanak (2020). Propositions d’écriture du iaai, langue parlée a Ouvéa, Nouvelle-Calédonie. Hna setr hwen iaai ae thep ûnyi. Académie des Langues Kanak. Bédi, B., ChatGPT-4 C-LARA-Instance, Chiera, B., Chua, C., Cucchiarini, C., Dotte, A.-L., Geneix-Rabault, Stéphanie Maizonniaux, C., M˘arginean, C., Ní Chiaráin, N., Parry-Mills, L., Raheb, C., Rayner, M., Simonsen, A., Viorica, Wacalie, F., Lucret,ia, M., Welby, P., Xiang, Z., and Zviel-Girshin, R. (2024). ChatGPT-Based Learning And Reading Assistant (C-LARA): Second report. Technical report. https://www.researchgate.net/publication/379119435_ChatGPT-Based_Learning_And_Reading_Assistant_C-LARA_Second_Report. Bédi, B., ChatGPT-4 C-LARA-Instance, Chiera, B., Chua, C., Cucchiarini, C., Ní Chiaráin, N., Rayner, M., Simonsen, A., and Zviel-Girshin, R. (2023). ChatGPT-Based Learning And Reading Assistant: Initial report. Technical report. https://www.researchgate.net/publication/372526096_ChatGPT-Based_Learning_And_Reading_Assistant_Initial_Report. Hombouy, M., Wea, G., and Goulon, I. (2005). Wanakat kaori — L’enfant kaori. ADCK-CCT-Grain de sable, Nouméa. Sam, L. D. (2009). Aqane cinyihanyi.

Fichier non déposé

Dates et versions

hal-05171752 , version 1 (21-07-2025)

Identifiants

  • HAL Id : hal-05171752 , version 1

Citer

Wejë Diane Bae, Chatgpt-Four C-Lara-Instance, Belinda Chiera, Anne-Laure Dotte, Stéphanie Geneix-Rabault, et al.. Creating Multimedia Resources for Kanak Languages Using the C-LARA Platform: Case Studies in Iaai and Drehu. COOL 13 - Conference on Oceanic Linguistics, ANU, Jun 2025, Canberra, Australia. ⟨hal-05171752⟩
196 Consultations
0 Téléchargements

Partager

  • More