<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="basu-2026-neural">
    <titleInfo>
        <title>Neural Nexus at PsyDefDetect: Fine-Tuning RoBERTa with Focal Loss and Role-Tagged Dialogue History for Defense Level Detection</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Subhrajyoti</namePart>
        <namePart type="family">Basu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the BioNLP 2026 (Shared Tasks)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Deepak</namePart>
            <namePart type="family">Gupta</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Dina</namePart>
            <namePart type="family">Demner-Fushman</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, California, USA</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-435-4</identifier>
    </relatedItem>
    <abstract>We describe our system for the PsyDefDetect shared task at BioNLP 2026, which focuses onclassifying help-seeker utterances in multi-turn supportive conversations into nine psychological defense mechanism levels defined by the Defense Mechanism Rating Scales (DMRS). Our approach fine-tunes roberta-base using a composite training objective that combines focal loss, label smoothing, and squareroot dampened class weights to address the severe label imbalance present in the PSYDEFCONV corpus, where the dominant class constitutes 52% of the training data. The inputrepresentation is constructed by concatenating up to eight dialogue turns with role-specific tags, separated using RoBERTa’s native /s tokens, followed by the target utterance marked using a [TARGET] token. Model selection is performed using macro-F1 based early stopping on a stratified 15% validation split, along with cosine learning rate decay for stable optimization. Our best submission achieves an official Leaderboard 1 (positive classes) macroF1 score of 0.2556, ranking 11th among 21 registered teams.</abstract>
    <identifier type="citekey">basu-2026-neural</identifier>
    <location>
        <url>https://aclanthology.org/2026.bionlp-2.10/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>66</start>
            <end>70</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Neural Nexus at PsyDefDetect: Fine-Tuning RoBERTa with Focal Loss and Role-Tagged Dialogue History for Defense Level Detection
%A Basu, Subhrajyoti
%Y Gupta, Deepak
%Y Demner-Fushman, Dina
%S Proceedings of the BioNLP 2026 (Shared Tasks)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-435-4
%F basu-2026-neural
%X We describe our system for the PsyDefDetect shared task at BioNLP 2026, which focuses onclassifying help-seeker utterances in multi-turn supportive conversations into nine psychological defense mechanism levels defined by the Defense Mechanism Rating Scales (DMRS). Our approach fine-tunes roberta-base using a composite training objective that combines focal loss, label smoothing, and squareroot dampened class weights to address the severe label imbalance present in the PSYDEFCONV corpus, where the dominant class constitutes 52% of the training data. The inputrepresentation is constructed by concatenating up to eight dialogue turns with role-specific tags, separated using RoBERTa’s native /s tokens, followed by the target utterance marked using a [TARGET] token. Model selection is performed using macro-F1 based early stopping on a stratified 15% validation split, along with cosine learning rate decay for stable optimization. Our best submission achieves an official Leaderboard 1 (positive classes) macroF1 score of 0.2556, ranking 11th among 21 registered teams.
%U https://aclanthology.org/2026.bionlp-2.10/
%P 66-70

Markdown (Informal)

[Neural Nexus at PsyDefDetect: Fine-Tuning RoBERTa with Focal Loss and Role-Tagged Dialogue History for Defense Level Detection](https://aclanthology.org/2026.bionlp-2.10/) (Basu, BioNLP 2026)

• Neural Nexus at PsyDefDetect: Fine-Tuning RoBERTa with Focal Loss and Role-Tagged Dialogue History for Defense Level Detection (Basu, BioNLP 2026)

ACL

• Subhrajyoti Basu. 2026. Neural Nexus at PsyDefDetect: Fine-Tuning RoBERTa with Focal Loss and Role-Tagged Dialogue History for Defense Level Detection. In Proceedings of the BioNLP 2026 (Shared Tasks), pages 66–70, San Diego, California, USA. Association for Computational Linguistics.