37.257, Software: Garo ASR - Speech to Text AI Model for Garo Language (A'chik)

[The LINGUIST List]

LINGUIST List: Vol-37-257. Mon Jan 19 2026. ISSN: 1069 - 4875.

Subject: 37.257, Software: Garo ASR - Speech to Text AI Model for Garo Language (A'chik)

Moderator: Steven Moran (linguist at linguistlist.org)
Managing Editor: Valeriia Vyshnevetska
Team: Helen Aristar-Dry, Mara Baccaro, Daniel Swanson
Jobs: jobs at linguistlist.org | Conferences: callconf at linguistlist.org | Pubs: pubs at linguistlist.org

Homepage: http://linguistlist.org

Editor for this issue: Daniel Swanson <daniel at linguistlist.org>

================================================================


Date: 18-Jan-2026
From: Badal Nyalang [nyalang at mwirelabs.com]
Subject: Garo ASR - Speech to Text AI Model for Garo Language (A'chik)


Garo ASR is an AI-powered automatic speech recognition system that
converts spoken Garo language (A'chik) into written text. Built using
transformer-based neural networks (Whisper architecture with 244M
parameters), the model achieves production quality transcription with
transfer learning approaches.
The system handles Latin-script Garo orthography used across
Meghalaya, Assam, and Tripura in India. Technical performance includes
9.74% Word Error Rate, 3.82% Character Error Rate on held-out test
data. The model processes various audio formats and operates in batch
mode for transcription tasks.
Key applications include language documentation, educational
technology, accessibility tools, media transcription, and cultural
preservation projects. The model is particularly suited for linguists
working on Tibeto-Burman languages, researchers documenting oral
traditions, and developers building language technology for Northeast
Indian languages.
Language(s): Garo (ISO 639-3: grt)
Script: Latin
Type of Software: Automatic Speech Recognition (ASR) / Speech-to-Text
Platform/Operating System: Cross-platform (Linux, macOS, Windows)
Technical Requirements:
Python 3.10+
PyTorch
Transformers library
Minimum 2GB RAM (8GB recommended)
GPU optional but recommended
Cost: Free
Developer/Organization: MWire Labs
Location: Shillong, Meghalaya, India
Website: https://mwirelabs.com/models/garo-asr/
Download/Access: https://huggingface.co/MWirelabs/garo-asr
Demo: https://huggingface.co/spaces/MWirelabs/garo-speech-to-text
Documentation: Available on HuggingFace model page
Version: 1.0
Release Date: January 2026

Linguistic Field(s): Applied Linguistics
                     Computational Linguistics

Subject Language(s): Garo (grt)

Language Family(ies): Sino-Tibetan
                      Tibeto-Burman



------------------------------------------------------------------------------

********************** LINGUIST List Support ***********************
Please consider donating to the Linguist List, a U.S. 501(c)(3) not for profit organization:

https://www.paypal.com/donate/?hosted_button_id=87C2AXTVC4PP8

LINGUIST List is supported by the following publishers:

Bloomsbury Publishing http://www.bloomsbury.com/uk/

Cambridge University Press http://www.cambridge.org/linguistics

Cascadilla Press http://www.cascadilla.com/

De Gruyter Brill https://www.degruyterbrill.com/?changeLang=en

Edinburgh University Press http://www.edinburghuniversitypress.com

John Benjamins http://www.benjamins.com/

Language Science Press http://langsci-press.org

Lincom GmbH https://lincom-shop.eu/

MIT Press http://mitpress.mit.edu/

Multilingual Matters http://www.multilingual-matters.com/

Narr Francke Attempto Verlag GmbH + Co. KG http://www.narr.de/

Netherlands Graduate School of Linguistics / Landelijke (LOT) http://www.lotpublications.nl/

Peter Lang AG http://www.peterlang.com


----------------------------------------------------------
LINGUIST List: Vol-37-257
----------------------------------------------------------