@inproceedings{huidrom-etal-2026-panini,
title = "{PANINI}: Improving Low-Resource {E}nglish-{M}anipuri Translation in Two Scripts via rs{L}o{RA} and Three-Stream Self-Augmentation",
author = "Huidrom, Rudali and
Kumar, Vikas and
Pangsatabam, Hoomexsun and
Das, Pinaki and
Khanganba, K. Kabi and
Zeno, Nongmaithem and
Khumukcham, Nicholas and
Okram, Mangalton and
Konjengbam, Justice and
Jamalpoor, Sai and
Khangembam, Alex and
Konjengbam, Anand and
Goyal, Vikram",
editor = "Haddow, Barry and
Kocmi, Tom and
Koehn, Philipp and
Monz, Christof",
booktitle = "Proceedings of the Eleventh Conference on Machine Translation",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.wmt-main.154/",
pages = "2154--2162",
ISBN = "979-8-89176-474-3",
abstract = "We describe team PANINI{'}s submission to the WMT 2026 Shared Task on Low-Resource Indic Language Translation. We submit primary systems for English{\ensuremath{\leftrightarrow}}Manipuri (Meiteilon) in both the Bengali (mni Beng) and Meitei Mayek (mni Mtei) scripts, covering all four directions. Rather than one shared multilingual adapter, we train four independent rank-stabilized LoRA (rsLoRA) adapters over IndicTrans2-1B, one per direction. Each adapter is trained on the authentic bitext plus three filtered synthetic streams the model generates for itself: back-translation, forward translation, and iterative pseudo-labelling, with subword regularisation and a script-aware processing pipeline. Our system obtains a COMET score of 92.31 and a BERTScore of 98.85 on English{\ensuremath{\leftrightarrow}}Manipuri (Meitei Mayek), the best in the direction. On the into-English directions, it reaches a COMET of 82.90 for mni Beng{\ensuremath{\leftrightarrow}}en and 80.20 for mniMtei{\ensuremath{\leftrightarrow}}en. These results show that per-direction adapters combined with self-generated synthetic data are effective even when authentic bitext is limited to a few thousand sentence pairs, and that a script-aware processing pipeline is necessary to produce well-formed output in both Manipuri scripts. The code and resources are released on GitHub."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huidrom-etal-2026-panini">
<titleInfo>
<title>PANINI: Improving Low-Resource English-Manipuri Translation in Two Scripts via rsLoRA and Three-Stream Self-Augmentation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rudali</namePart>
<namePart type="family">Huidrom</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vikas</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hoomexsun</namePart>
<namePart type="family">Pangsatabam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pinaki</namePart>
<namePart type="family">Das</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">K</namePart>
<namePart type="given">Kabi</namePart>
<namePart type="family">Khanganba</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nongmaithem</namePart>
<namePart type="family">Zeno</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nicholas</namePart>
<namePart type="family">Khumukcham</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mangalton</namePart>
<namePart type="family">Okram</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Justice</namePart>
<namePart type="family">Konjengbam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sai</namePart>
<namePart type="family">Jamalpoor</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alex</namePart>
<namePart type="family">Khangembam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anand</namePart>
<namePart type="family">Konjengbam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vikram</namePart>
<namePart type="family">Goyal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Eleventh Conference on Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Barry</namePart>
<namePart type="family">Haddow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tom</namePart>
<namePart type="family">Kocmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philipp</namePart>
<namePart type="family">Koehn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Monz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-474-3</identifier>
</relatedItem>
<abstract>We describe team PANINI’s submission to the WMT 2026 Shared Task on Low-Resource Indic Language Translation. We submit primary systems for English\ensuremathłeftrightarrowManipuri (Meiteilon) in both the Bengali (mni Beng) and Meitei Mayek (mni Mtei) scripts, covering all four directions. Rather than one shared multilingual adapter, we train four independent rank-stabilized LoRA (rsLoRA) adapters over IndicTrans2-1B, one per direction. Each adapter is trained on the authentic bitext plus three filtered synthetic streams the model generates for itself: back-translation, forward translation, and iterative pseudo-labelling, with subword regularisation and a script-aware processing pipeline. Our system obtains a COMET score of 92.31 and a BERTScore of 98.85 on English\ensuremathłeftrightarrowManipuri (Meitei Mayek), the best in the direction. On the into-English directions, it reaches a COMET of 82.90 for mni Beng\ensuremathłeftrightarrowen and 80.20 for mniMtei\ensuremathłeftrightarrowen. These results show that per-direction adapters combined with self-generated synthetic data are effective even when authentic bitext is limited to a few thousand sentence pairs, and that a script-aware processing pipeline is necessary to produce well-formed output in both Manipuri scripts. The code and resources are released on GitHub.</abstract>
<identifier type="citekey">huidrom-etal-2026-panini</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.154/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>2154</start>
<end>2162</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PANINI: Improving Low-Resource English-Manipuri Translation in Two Scripts via rsLoRA and Three-Stream Self-Augmentation
%A Huidrom, Rudali
%A Kumar, Vikas
%A Pangsatabam, Hoomexsun
%A Das, Pinaki
%A Khanganba, K. Kabi
%A Zeno, Nongmaithem
%A Khumukcham, Nicholas
%A Okram, Mangalton
%A Konjengbam, Justice
%A Jamalpoor, Sai
%A Khangembam, Alex
%A Konjengbam, Anand
%A Goyal, Vikram
%Y Haddow, Barry
%Y Kocmi, Tom
%Y Koehn, Philipp
%Y Monz, Christof
%S Proceedings of the Eleventh Conference on Machine Translation
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-474-3
%F huidrom-etal-2026-panini
%X We describe team PANINI’s submission to the WMT 2026 Shared Task on Low-Resource Indic Language Translation. We submit primary systems for English\ensuremathłeftrightarrowManipuri (Meiteilon) in both the Bengali (mni Beng) and Meitei Mayek (mni Mtei) scripts, covering all four directions. Rather than one shared multilingual adapter, we train four independent rank-stabilized LoRA (rsLoRA) adapters over IndicTrans2-1B, one per direction. Each adapter is trained on the authentic bitext plus three filtered synthetic streams the model generates for itself: back-translation, forward translation, and iterative pseudo-labelling, with subword regularisation and a script-aware processing pipeline. Our system obtains a COMET score of 92.31 and a BERTScore of 98.85 on English\ensuremathłeftrightarrowManipuri (Meitei Mayek), the best in the direction. On the into-English directions, it reaches a COMET of 82.90 for mni Beng\ensuremathłeftrightarrowen and 80.20 for mniMtei\ensuremathłeftrightarrowen. These results show that per-direction adapters combined with self-generated synthetic data are effective even when authentic bitext is limited to a few thousand sentence pairs, and that a script-aware processing pipeline is necessary to produce well-formed output in both Manipuri scripts. The code and resources are released on GitHub.
%U https://aclanthology.org/2026.wmt-main.154/
%P 2154-2162
Markdown (Informal)
[PANINI: Improving Low-Resource English-Manipuri Translation in Two Scripts via rsLoRA and Three-Stream Self-Augmentation](https://aclanthology.org/2026.wmt-main.154/) (Huidrom et al., WMT 2026)
ACL
- Rudali Huidrom, Vikas Kumar, Hoomexsun Pangsatabam, Pinaki Das, K. Kabi Khanganba, Nongmaithem Zeno, Nicholas Khumukcham, Mangalton Okram, Justice Konjengbam, Sai Jamalpoor, Alex Khangembam, Anand Konjengbam, and Vikram Goyal. 2026. PANINI: Improving Low-Resource English-Manipuri Translation in Two Scripts via rsLoRA and Three-Stream Self-Augmentation. In Proceedings of the Eleventh Conference on Machine Translation, pages 2154–2162, Budapest, Hungary. Association for Computational Linguistics.