@inproceedings{jagadeeshan-etal-2026-padamitra,
title = "Padamitra: Grounded Glossary Generation for Classical {S}anskrit",
author = "Jagadeeshan, Manoj Balaji and
Marala, Sai Pragnaan and
Goyal, Pawan",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2007/",
doi = "10.18653/2026.findings-emnlp.2007",
pages = "42745--42755",
ISBN = "979-8-89176-462-0",
abstract = "We introduce grounded glossary generation, a structured task requiring models to recover semantically meaningful Sanskrit phrases and produce translation-grounded meanings from a {\'s}loka-translation pair, formalizing the traditional p{\={a}}ṭha-commentary practice as an evaluable NLP objective. We construct a benchmark of 31,316 {\'s}loka-translation-glossary triples from the V{\={a}}lm{\={i}}ki R{\={a}}m{\={a}}yaṇa and {\'S}r{\={i}}mad Bh{\={a}}gavatam, paired with two metrics: Jaccard for phrase recovery and Meaning Faithfulness for semantic consistency. Across zero-shot, few-shot, and instruction fine-tuned variants of Gemma-3n-E4B, Gemma-3-12B, Phi-4, and Qwen3.5-9B, instruction fine-tuning substantially outperforms prompting, while explicit segmentation yields gains. Error analysis identifies over-segmentation of sandhi and sam{\={a}}sa compounds as the dominant failure mode, pointing to morphological modeling as the key bottleneck for faithful Sanskrit lexical decomposition."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jagadeeshan-etal-2026-padamitra">
<titleInfo>
<title>Padamitra: Grounded Glossary Generation for Classical Sanskrit</title>
</titleInfo>
<name type="personal">
<namePart type="given">Manoj</namePart>
<namePart type="given">Balaji</namePart>
<namePart type="family">Jagadeeshan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sai</namePart>
<namePart type="given">Pragnaan</namePart>
<namePart type="family">Marala</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pawan</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>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>We introduce grounded glossary generation, a structured task requiring models to recover semantically meaningful Sanskrit phrases and produce translation-grounded meanings from a śloka-translation pair, formalizing the traditional pāṭha-commentary practice as an evaluable NLP objective. We construct a benchmark of 31,316 śloka-translation-glossary triples from the Vālmīki Rāmāyaṇa and Śrīmad Bhāgavatam, paired with two metrics: Jaccard for phrase recovery and Meaning Faithfulness for semantic consistency. Across zero-shot, few-shot, and instruction fine-tuned variants of Gemma-3n-E4B, Gemma-3-12B, Phi-4, and Qwen3.5-9B, instruction fine-tuning substantially outperforms prompting, while explicit segmentation yields gains. Error analysis identifies over-segmentation of sandhi and samāsa compounds as the dominant failure mode, pointing to morphological modeling as the key bottleneck for faithful Sanskrit lexical decomposition.</abstract>
<identifier type="citekey">jagadeeshan-etal-2026-padamitra</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2007</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2007/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>42745</start>
<end>42755</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Padamitra: Grounded Glossary Generation for Classical Sanskrit
%A Jagadeeshan, Manoj Balaji
%A Marala, Sai Pragnaan
%A Goyal, Pawan
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F jagadeeshan-etal-2026-padamitra
%X We introduce grounded glossary generation, a structured task requiring models to recover semantically meaningful Sanskrit phrases and produce translation-grounded meanings from a śloka-translation pair, formalizing the traditional pāṭha-commentary practice as an evaluable NLP objective. We construct a benchmark of 31,316 śloka-translation-glossary triples from the Vālmīki Rāmāyaṇa and Śrīmad Bhāgavatam, paired with two metrics: Jaccard for phrase recovery and Meaning Faithfulness for semantic consistency. Across zero-shot, few-shot, and instruction fine-tuned variants of Gemma-3n-E4B, Gemma-3-12B, Phi-4, and Qwen3.5-9B, instruction fine-tuning substantially outperforms prompting, while explicit segmentation yields gains. Error analysis identifies over-segmentation of sandhi and samāsa compounds as the dominant failure mode, pointing to morphological modeling as the key bottleneck for faithful Sanskrit lexical decomposition.
%R 10.18653/2026.findings-emnlp.2007
%U https://aclanthology.org/2026.findings-emnlp.2007/
%U https://doi.org/10.18653/2026.findings-emnlp.2007
%P 42745-42755
Markdown (Informal)
[Padamitra: Grounded Glossary Generation for Classical Sanskrit](https://aclanthology.org/2026.findings-emnlp.2007/) (Jagadeeshan et al., Findings 2026)
ACL