<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="seo-etal-2026-domain">
    <titleInfo>
        <title>Domain-Agnostic Neural Topic Modeling with Contextual Token-Level Semantic Graph Representation</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Seung-Won</namePart>
        <namePart type="family">Seo</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Won</namePart>
        <namePart type="given">Ik</namePart>
        <namePart type="family">Cho</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yongmin</namePart>
        <namePart type="family">Yoo</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-11</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Yue</namePart>
            <namePart type="family">Zhang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ruifeng</namePart>
            <namePart type="family">Xu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ayu</namePart>
            <namePart type="family">Purwarianti</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Tong</namePart>
            <namePart type="family">Rong</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jun</namePart>
            <namePart type="family">Suzuki</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Hengqin, China</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-439-2</identifier>
    </relatedItem>
    <abstract>PLM-powered neural topic models depend on general-domain pre-training, leading to degraded topic interpretability on specialized corpora. The root cause is geometric: domain-specific terms unseen during pre-training collapse into an indistinguishable region of the embedding space, and neither domain-specific re-training, word-level graph enrichment, nor parameter-efficient fine-tuning can restructure this space without inheriting the capacity ceiling of the underlying encoder. Our key insight is that a learnable graph layer operating on token-level PLM embeddings can acquire corpus-specific semantic structure that the frozen encoder lacks, because token-level graphs preserve document-local context that word-level representations discard and joint optimization with the topic objective reshapes embedding geometry directly from target-domain evidence. We instantiate this insight as DARTopic, a domain-agnostic framework that constructs token-level semantic graphs from frozen PLM embeddings and jointly trains a GNN encoder with variational topic inference. Across three benchmarks spanning general, biomedical, and legal domains, DARTopic consistently outperforms strong baselines in topic coherence and document clustering without any encoder fine-tuning, while demonstrating robustness to PLM choice and favorable runtime efficiency over fine-tuning-based alternatives.</abstract>
    <identifier type="citekey">seo-etal-2026-domain</identifier>
    <location>
        <url>https://aclanthology.org/2026.aacl-long.241/</url>
    </location>
    <part>
        <date>2026-11</date>
        <extent unit="page">
            <start>5415</start>
            <end>5429</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Domain-Agnostic Neural Topic Modeling with Contextual Token-Level Semantic Graph Representation
%A Seo, Seung-Won
%A Cho, Won Ik
%A Yoo, Yongmin
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F seo-etal-2026-domain
%X PLM-powered neural topic models depend on general-domain pre-training, leading to degraded topic interpretability on specialized corpora. The root cause is geometric: domain-specific terms unseen during pre-training collapse into an indistinguishable region of the embedding space, and neither domain-specific re-training, word-level graph enrichment, nor parameter-efficient fine-tuning can restructure this space without inheriting the capacity ceiling of the underlying encoder. Our key insight is that a learnable graph layer operating on token-level PLM embeddings can acquire corpus-specific semantic structure that the frozen encoder lacks, because token-level graphs preserve document-local context that word-level representations discard and joint optimization with the topic objective reshapes embedding geometry directly from target-domain evidence. We instantiate this insight as DARTopic, a domain-agnostic framework that constructs token-level semantic graphs from frozen PLM embeddings and jointly trains a GNN encoder with variational topic inference. Across three benchmarks spanning general, biomedical, and legal domains, DARTopic consistently outperforms strong baselines in topic coherence and document clustering without any encoder fine-tuning, while demonstrating robustness to PLM choice and favorable runtime efficiency over fine-tuning-based alternatives.
%U https://aclanthology.org/2026.aacl-long.241/
%P 5415-5429

Markdown (Informal)

[Domain-Agnostic Neural Topic Modeling with Contextual Token-Level Semantic Graph Representation](https://aclanthology.org/2026.aacl-long.241/) (Seo et al., AACL-IJCNLP 2026)

• Domain-Agnostic Neural Topic Modeling with Contextual Token-Level Semantic Graph Representation (Seo et al., AACL-IJCNLP 2026)

ACL

• Seung-Won Seo, Won Ik Cho, and Yongmin Yoo. 2026. Domain-Agnostic Neural Topic Modeling with Contextual Token-Level Semantic Graph Representation. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5415–5429, Hengqin, China. Association for Computational Linguistics.