@inproceedings{saikia-etal-2026-morphological,
title = "Morphological Analysis and Disambiguation of Agglutinative {A}ssamese through Rule-Based Synthesis and Neural Candidate Ranking",
author = "Saikia, Pragyanur and
Sharma, Utpal and
Karmakar, Arindam",
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.2118/",
doi = "10.18653/2026.findings-emnlp.2118",
pages = "45155--45167",
ISBN = "979-8-89176-462-0",
abstract = "Assamese is a highly agglutinative language that uses suffix sequences to convey various kinds of information. A single morpheme may exhibit multiple syntactic or semantic functions depending on its position in a word and the preceding morpheme. Such ambiguities and the inadequacy of supervised data make morphological analysis difficult for Assamese. Our approach is based on the observation that a word has a limited set of valid morphological analyses, among which one is contextually the most appropriate. We curate a morphological lexicon, a bigram transition rule set, and several derivational rules, which our rule-based analyzer uses to generate valid analyses of a word by segmenting its surface form. The rule-based analyzer facilitates context-based annotation of over 11.8K tokens across more than 1K sentences. We construct morphological feature vectors from each analysis of a word and train a neural model to discriminate between correct and incorrect sequences of morphological vectors. The trained model is evaluated on the task of assigning the highest score to the correct sequence of morphological vectors among all the candidate sequences. The model achieves a word-level disambiguation accuracy of 80.76{\%} by resolving 1045 out of 1294 ambiguities on the test set without using lexical semantics. The results indicate regularity in sentence-level morphological patterns. The annotated dataset and other resources are available at https://github.com/Pragyanur/morphological-analysis-assamese."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="saikia-etal-2026-morphological">
<titleInfo>
<title>Morphological Analysis and Disambiguation of Agglutinative Assamese through Rule-Based Synthesis and Neural Candidate Ranking</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pragyanur</namePart>
<namePart type="family">Saikia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Utpal</namePart>
<namePart type="family">Sharma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arindam</namePart>
<namePart type="family">Karmakar</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>Assamese is a highly agglutinative language that uses suffix sequences to convey various kinds of information. A single morpheme may exhibit multiple syntactic or semantic functions depending on its position in a word and the preceding morpheme. Such ambiguities and the inadequacy of supervised data make morphological analysis difficult for Assamese. Our approach is based on the observation that a word has a limited set of valid morphological analyses, among which one is contextually the most appropriate. We curate a morphological lexicon, a bigram transition rule set, and several derivational rules, which our rule-based analyzer uses to generate valid analyses of a word by segmenting its surface form. The rule-based analyzer facilitates context-based annotation of over 11.8K tokens across more than 1K sentences. We construct morphological feature vectors from each analysis of a word and train a neural model to discriminate between correct and incorrect sequences of morphological vectors. The trained model is evaluated on the task of assigning the highest score to the correct sequence of morphological vectors among all the candidate sequences. The model achieves a word-level disambiguation accuracy of 80.76% by resolving 1045 out of 1294 ambiguities on the test set without using lexical semantics. The results indicate regularity in sentence-level morphological patterns. The annotated dataset and other resources are available at https://github.com/Pragyanur/morphological-analysis-assamese.</abstract>
<identifier type="citekey">saikia-etal-2026-morphological</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2118</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2118/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>45155</start>
<end>45167</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Morphological Analysis and Disambiguation of Agglutinative Assamese through Rule-Based Synthesis and Neural Candidate Ranking
%A Saikia, Pragyanur
%A Sharma, Utpal
%A Karmakar, Arindam
%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 saikia-etal-2026-morphological
%X Assamese is a highly agglutinative language that uses suffix sequences to convey various kinds of information. A single morpheme may exhibit multiple syntactic or semantic functions depending on its position in a word and the preceding morpheme. Such ambiguities and the inadequacy of supervised data make morphological analysis difficult for Assamese. Our approach is based on the observation that a word has a limited set of valid morphological analyses, among which one is contextually the most appropriate. We curate a morphological lexicon, a bigram transition rule set, and several derivational rules, which our rule-based analyzer uses to generate valid analyses of a word by segmenting its surface form. The rule-based analyzer facilitates context-based annotation of over 11.8K tokens across more than 1K sentences. We construct morphological feature vectors from each analysis of a word and train a neural model to discriminate between correct and incorrect sequences of morphological vectors. The trained model is evaluated on the task of assigning the highest score to the correct sequence of morphological vectors among all the candidate sequences. The model achieves a word-level disambiguation accuracy of 80.76% by resolving 1045 out of 1294 ambiguities on the test set without using lexical semantics. The results indicate regularity in sentence-level morphological patterns. The annotated dataset and other resources are available at https://github.com/Pragyanur/morphological-analysis-assamese.
%R 10.18653/2026.findings-emnlp.2118
%U https://aclanthology.org/2026.findings-emnlp.2118/
%U https://doi.org/10.18653/2026.findings-emnlp.2118
%P 45155-45167
Markdown (Informal)
[Morphological Analysis and Disambiguation of Agglutinative Assamese through Rule-Based Synthesis and Neural Candidate Ranking](https://aclanthology.org/2026.findings-emnlp.2118/) (Saikia et al., Findings 2026)
ACL