# Low-resource Language Support

A speech-to-text translation model for Indian languages fine-tuned on SeamlessM4T-v2, supporting 13 Indian languages with performance surpassing the base model and competing systems.

NLLB-200-3.3B is a multilingual neural machine translation model developed by Meta (formerly Facebook), supporting translation tasks between 200 languages.

OWLS is a suite of Whisper-style models designed to help researchers understand the scaling properties of speech models, supporting multilingual speech recognition and translation.

A text-to-speech model supporting Zaghawa and Siri languages, suitable for speech synthesis applications.

This is a multilingual translation model that supports translation from multiple languages to multiple languages. It is trained on a large amount of public data and can be used for translation and text generation tasks.

English to Nepali translation model fine-tuned on unknown datasets based on the NLLB-200 distilled model

bge-fr-en is a sentence embedding model based on sentence-transformers, specializing in sentence similarity calculation and feature extraction for French and English.

The Shivneri Marathi LLM is designed to provide generative AI capabilities for non-English (especially Marathi) users in India, supporting text generation in both Marathi and English.

A fine-tuned Transformer-based sequence-to-sequence translation model for English to Tamil

Bemba (bem) text-to-speech model developed by Meta, part of the Massively Multilingual Speech project

This is a model for Marathi speech processing, potentially involving speech recognition or speech synthesis tasks.

mREBEL is the multilingual version of REBEL for multilingual relation extraction tasks, supporting 18 languages.

A named entity recognition model for 21 African languages, fine-tuned based on the Davlan/afro-xlmr-large model, supporting the recognition of four entity types: dates, locations, organizations, and person names.

MT0-XL is a multilingual text generation model that supports multiple languages and tasks, including coreference resolution and natural language inference.

This is a pre-trained language model specifically for Nepali, with research results published in the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing (2022).

BLOOMZ-3B1 is a multilingual generative model from the BLOOM series, supporting 46 languages and 13 programming languages, excelling in cross-lingual text generation tasks.

BLOOMZ-7B1 is a multilingual, multitask large-scale language model supporting over 40 languages and multiple programming languages, suitable for various natural language processing tasks such as text generation and sentiment analysis.

This model is specifically designed for the SemEval Multiconer task, serving as a named entity recognition (NER) model to identify complex entity categories in multilingual and cross-domain texts.

A large-scale multilingual speech recognition model introduced by Meta AI, supporting 60 languages, based on a 1-billion-parameter Transformer encoder architecture.

AfroXLMR-large is a multilingual model created by performing masked language modeling adaptation training on the XLM-R-large model for 17 African languages.

CINO is a multilingual pretrained model for Chinese minority languages, supporting Chinese and 7 minority languages.

A multilingual pretrained model designed for Chinese minority languages, supporting Chinese and 7 minority languages.

An automatic speech recognition model fine-tuned on Persian (Farsi) using the Common Voice dataset, based on facebook/wav2vec2-large-xlsr-53

Fine-tuned based on Facebook's wav2vec2-large-xlsr-53 model, supporting automatic speech recognition tasks for Hindi and Marathi

A Transformer-based machine translation model from English to Tok Pisin, developed by the Helsinki-NLP team.

opus-mt-de-kg is a machine translation model based on the Transformer architecture, specifically designed for translating German (de) to Kikongo (kg).

This is a machine translation model from English to Irish based on the transformer-align architecture, released by the Tatoeba-Challenge team.

This is a machine translation model based on the Transformer architecture, designed to translate English (en) to Kwangali (kwn) language.

This is a Transformer-based machine translation model for Spanish to Xhosa, developed by the Helsinki-NLP team.

A Transformer-based machine translation model for Spanish to Igbo, developed by the Helsinki-NLP team and trained using the OPUS dataset.

This is a machine translation model based on the Transformer architecture, designed for translating English to Kinyarwanda.

opus-mt-en-lg is a machine translation model based on the Transformer architecture, designed to translate English (en) to Luganda (lg). This model was developed and released by the Helsinki University Natural Language Processing team (Helsinki-NLP).

A machine translation model trained on the OPUS dataset for English to Kwy language translation, using transformer-align architecture

This is a machine translation model based on the Transformer architecture for translating English to Luo language.

OPUS-MT machine translation model for Ewe (ee) to German (de), based on Transformer architecture and SentencePiece preprocessing.

opus-mt-fr-tum is a machine translation model based on the Transformer architecture, specifically designed for translation tasks from French (fr) to Tumbuka (tum).

This is a Transformer-based machine translation model specifically designed for translating French into Tahitian.

opus-mt-fr-lg is a machine translation model based on the transformer-align architecture, supporting translation tasks from French (fr) to Luganda (lg).

opus-mt-fr-ln is a machine translation model based on the transformer-align architecture, specifically designed for translating French (fr) to Lingala (ln). This model was developed by the Helsinki-NLP team and trained using the OPUS dataset.

A Transformer-based machine translation model for French to Sango, developed by the Helsinki-NLP team.