# Low-Resource Language Optimization

EMMA-500 is a multilingual large language model based on the Llama 2 7B architecture, supporting over 500 languages and excelling in tasks such as commonsense reasoning and machine translation.

This model is a fine-tuned English to Dzongkha translation model based on facebook/nllb-200-distilled-600M

The first high-performance open-source language model for Wolof, built on the Qwen 2.5 architecture, supporting bidirectional translation between Wolof and English, text generation, and other tasks

A multilingual processing model supporting over 150 languages, suitable for various natural language processing tasks.

NusaBERT Base Version is a multilingual encoder language model based on the BERT architecture, supporting 13 Indonesian regional languages and pretrained on multiple open-source corpora.

AfroXLMR-large-76L is created through MLM (Masked Language Modeling) adaptation of the extended XLM-R-large model, covering 76 widely used languages in Africa, including 4 high-resource languages.

A Myanmar language model based on the GPT2 architecture, supporting Burmese text understanding and generation

This is a general-purpose model supporting over 150 languages, suitable for various natural language processing tasks.

NLLB-200-3.3B is a large-scale multilingual translation model supporting 200 languages, trained on the FLORES-200 dataset.

A multilingual processing model collection supporting over 100 languages and writing systems

A BERT-based masked language model for Buddhist Sanskrit text processing, specifically designed for handling Buddhist Sanskrit texts

This is a machine translation model based on the Transformer architecture, supporting translation from English to various Creole and Pidgin mixed languages, including Antillean Creole, Haitian Creole, and Mauritian Creole.

This is a multilingual translation model based on the Transformer architecture, supporting translation tasks from various Afro-Asiatic languages to English.

This is a Malay-to-Malay translation model based on the transformer-align architecture, supporting multiple Malay variants.

An English-Yoruba machine translation model fine-tuned on mbart-large-50, trained on JW300 and Menyo-20k datasets