LexiKa: a collaborative lexical database for Kanak languages
Résumé
While current language models (LLMs) are mainly trained on massively resourced dominant languages, creating a significant linguistic divide with so-called “rare” indigenous languages, the web-based dictionary management application developed by the DiKaLa project represents a significant innovation in the digital transition and enhancement of Pacific languages. Developed specifically for the Kanak languages of New Caledonia, this application enables collaborative editing of data extracted from digitized historical dictionaries, as well as their augmentation, while respecting the linguistic specificities of each language (phonetic inventory, orthographic choices, morpho-lexical structuring). The application combines modern technology with traditional knowledge, enabling linguists, researchers and native speakers to work together on existing dictionaries. It incorporates advanced functionalities including a modification traceability system to preserve the history of collaborative interventions; an intuitive interface for adding and modifying definitions, examples and metadata, as well as an option for validating or updating the orthography of entries in order to participate in the dissemination of official and recent writing proposals for these languages. Diatopic variations are also taken into account thanks to the possibility of Diatopic variations are also taken into account, thanks to the possibility of localizing dialectal variants with a place, district or area name. This application represents a first phase in the process of creating digital tools targeted at Kanak languages, and is intended to be transposable to other Oceanic languages. It is part of a perspective in which modern digital technologies can serve the preservation of the Oceanic linguistic heritage, thus contributing to the vitality and transmission of Pacific languages for future generations.