Findings of WMT 2024 Shared Task on Low-Resource Indic Languages Translation

This technical report presents the results of the low-resource Indic language translation task conducted during the Ninth Conference on Machine Translation (WMT) 2024. The task involved developing machine translation models for four language pairs: English–Assamese, English–Mizo, English–Khasi, and English–Manipuri, utilizing the IndicNE-Corp1.0 dataset, which comprises a rich collection of parallel and monolingual corpora for northeastern Indic languages. The evaluation of the translation systems was carried out using a comprehensive suite of automatic metrics, including BLEU, TER, RIBES, METEOR, and ChrF, along with detailed human assessments to measure performance and accuracy. The initiative aims to advance low-resource machine translation and contribute significantly to the field, addressing the challenges faced by low-resource Indic languages, which often lack sufficient resources and institutional support. The report emphasizes the importance of continued efforts in documenting and revitalizing these languages through technological solutions.