@article{ART003397381},
author={Myeong-Guk Jeong and Ji-Ho Kim and Dana Yang},
title={Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={10},
pages={71-81}
TY - JOUR
AU - Myeong-Guk Jeong
AU - Ji-Ho Kim
AU - Dana Yang
TI - Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 10
PB - The Korean Society Of Computer And Information
SP - 71
EP - 81
SN - 1598-849X
AB - Mental health text classification based on social media is limited by the scarcity of labeled data and the computational burden associated with full-parameter fine-tuning. To address these limitations, this paper proposes an Enhanced LoRA-SimMatch(ELSM) model that applies LoRA to a pre-trained language model and integrates it with SimMatch-based semi-supervised learning. Through this structure, the issue of pseudo-label noise in semi-supervised learning is effectively refined, enabling precise classification of mental health texts. In experimental results, while training only 1.05% of the total parameters, ELSM achieves relative improvements of 8.04% in Accuracy and 7.84% in Macro-F1, alongside a 36.3% reduction in ECE compared to existing baselines, demonstrating superior classification performance and prediction reliability. As future research, we plan to evaluate the domain extensibility of ELSM in low-resource labeled-data environments across various text processing tasks, including legal document classification, medical text analysis, and toxic text detection.
KW - Semi-Supervised Learning;LoRA;SimMatch;PLM;Mental Health Text Classification
DO -
UR -
ER -
Myeong-Guk Jeong, Ji-Ho Kim and Dana Yang. (2026). Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments. Journal of The Korea Society of Computer and Information, 31(10), 71-81.
Myeong-Guk Jeong, Ji-Ho Kim and Dana Yang. 2026, "Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments", Journal of The Korea Society of Computer and Information, vol.31, no.10 pp.71-81.
Myeong-Guk Jeong, Ji-Ho Kim, Dana Yang "Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments" Journal of The Korea Society of Computer and Information 31.10 pp.71-81 (2026) : 71.
Myeong-Guk Jeong, Ji-Ho Kim, Dana Yang. Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments. 2026; 31(10), 71-81.
Myeong-Guk Jeong, Ji-Ho Kim and Dana Yang. "Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments" Journal of The Korea Society of Computer and Information 31, no.10 (2026) : 71-81.
Myeong-Guk Jeong; Ji-Ho Kim; Dana Yang. Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments. Journal of The Korea Society of Computer and Information, 31(10), 71-81.
Myeong-Guk Jeong; Ji-Ho Kim; Dana Yang. Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments. Journal of The Korea Society of Computer and Information. 2026; 31(10) 71-81.
Myeong-Guk Jeong, Ji-Ho Kim, Dana Yang. Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments. 2026; 31(10), 71-81.
Myeong-Guk Jeong, Ji-Ho Kim and Dana Yang. "Enhanced LoRA-SimMatch Model for Mental Health-Related Text Classification in Social Media Environments" Journal of The Korea Society of Computer and Information 31, no.10 (2026) : 71-81.
1