@inproceedings{kobayashi-li-2026-greedy,
title = "Greedy Matching Aggregation for Soft Contrastive Learning in Multi-Label Text Classification",
author = "Kobayashi, Ryoma and
Li, Jiyi",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.62/",
pages = "1222--1235",
ISBN = "979-8-89176-444-6",
abstract = "Multi-label text classification is important for organizing large document collections, but correlated labels and skewed label distributions make representation learning difficult. Existing multi-label contrastive methods usually define pairwise supervision from label overlap, resulting in a coarse signal limited to observed shared labels. We propose Greedy Matching Aggregation Soft Contrastive Learning (GMA-SCL), which constructs contrastive targets for MLTC by greedily matching label texts and incorporating label importance weighting. The resulting reference distribution is sharpened to emphasize highly similar candidate documents while preserving graded supervision from moderately similar ones. Experiments on AAPD, Reuters-21578, and UK-LEX show that GMA-SCL improves Macro-F1 over existing multi-label contrastive baselines while maintaining competitive Micro-F1 and Hamming Loss. Our code is available at https://github.com/koba-84/gma-scl."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kobayashi-li-2026-greedy">
<titleInfo>
<title>Greedy Matching Aggregation for Soft Contrastive Learning in Multi-Label Text Classification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ryoma</namePart>
<namePart type="family">Kobayashi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiyi</namePart>
<namePart type="family">Li</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>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</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">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Multi-label text classification is important for organizing large document collections, but correlated labels and skewed label distributions make representation learning difficult. Existing multi-label contrastive methods usually define pairwise supervision from label overlap, resulting in a coarse signal limited to observed shared labels. We propose Greedy Matching Aggregation Soft Contrastive Learning (GMA-SCL), which constructs contrastive targets for MLTC by greedily matching label texts and incorporating label importance weighting. The resulting reference distribution is sharpened to emphasize highly similar candidate documents while preserving graded supervision from moderately similar ones. Experiments on AAPD, Reuters-21578, and UK-LEX show that GMA-SCL improves Macro-F1 over existing multi-label contrastive baselines while maintaining competitive Micro-F1 and Hamming Loss. Our code is available at https://github.com/koba-84/gma-scl.</abstract>
<identifier type="citekey">kobayashi-li-2026-greedy</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.62/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>1222</start>
<end>1235</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Greedy Matching Aggregation for Soft Contrastive Learning in Multi-Label Text Classification
%A Kobayashi, Ryoma
%A Li, Jiyi
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F kobayashi-li-2026-greedy
%X Multi-label text classification is important for organizing large document collections, but correlated labels and skewed label distributions make representation learning difficult. Existing multi-label contrastive methods usually define pairwise supervision from label overlap, resulting in a coarse signal limited to observed shared labels. We propose Greedy Matching Aggregation Soft Contrastive Learning (GMA-SCL), which constructs contrastive targets for MLTC by greedily matching label texts and incorporating label importance weighting. The resulting reference distribution is sharpened to emphasize highly similar candidate documents while preserving graded supervision from moderately similar ones. Experiments on AAPD, Reuters-21578, and UK-LEX show that GMA-SCL improves Macro-F1 over existing multi-label contrastive baselines while maintaining competitive Micro-F1 and Hamming Loss. Our code is available at https://github.com/koba-84/gma-scl.
%U https://aclanthology.org/2026.findings-aacl.62/
%P 1222-1235
Markdown (Informal)
[Greedy Matching Aggregation for Soft Contrastive Learning in Multi-Label Text Classification](https://aclanthology.org/2026.findings-aacl.62/) (Kobayashi & Li, Findings 2026)
ACL