Adaptive Fusion of DeBERTa-Base and MC-CNN for Enhanced Text Classification by Using AG News Dataset
DOI:
https://doi.org/10.29304/jqcsm.2026.18.33219Keywords:
Natural Language Processing (NLP), MC-CNN and DeBERTaAbstract
Text classification are an significant task in natural language processing(NLP). Different deep learning models can capture different types of information from text. This study proposes a new adaptive fusion method that combines MC-CNN and DeBERTa for text classification. MC-CNN are used to obtain local text features while DeBERTa captures contextual and semantic information. An adaptive fusion mechanism is proposed to dynamically combine the outputs for both models and improve classification performance. The presented model achieved the best results, with 95.62% accuracy, 95.01% precision, 94.38% recall, and 94.69% F1-score. The aim of this study is develop a simple and effective deep learning framework that combines the complementary strengths of CNN and Transformer-based models for improved text classification.
Downloads
References
REFERENCES
E. Boitel, A. Mohasseb, and E. Haig, “MIST : Multimodal emotion recognition using DeBERTa for text , Semi-CNN for speech , ResNet-50 for facial , and 3D-CNN for motion analysis,” Expert Syst. Appl., vol. 270, no. April 2024, p. 126236, 2025, doi: 10.1016/j.eswa.2024.126236.
E. Boitel, A. Mohasseb, and E. Haig, “MIST : Multimodal emotion recognition using DeBERTa for text , Semi-CNN for speech , ResNet-50 for facial , and 3D-CNN for motion analysis,” vol. 270, no. April 2024, 2025.
A. Assiri, A. Gumaei, F. Mehmood, T. Abbas, and S. Ullah, “DeBERTa-GRU : Sentiment Analysis for Large Language Model,” no. Ml, 2024, doi: 10.32604/cmc.2024.050781.
B. S. Al-smadi, “DeBERTa-BiLSTM : A multi-label classification model of Arabic medical questions using pre-trained models and deep learning,” Comput. Biol. Med., vol. 170, no. January, p. 107921, 2024, doi: 10.1016/j.compbiomed.2024.107921.
A. Seerat, S. Nasir, M. Wasim, and N. M. Garcia, “ScienceDirect Biomedical Biomedical Natural Natural Language Language Inference Inference on on Clinical Clinical trials trials using using the the Models BERT-based Models,” Procedia Comput. Sci., vol. 241, no. 2019, pp. 576–581, 2024, doi: 10.1016/j.procs.2024.08.083.
Q. Zhang, X. Xiang, Y. Li, and Y. Hou, “ScienceDirect Machine Learning and NLP Fusion Algorithms for Short Text Classification,” vol. 285, pp. 10–18, 2026, doi: 10.1016/j.procs.2026.07.165.
H. Black, R. Kurt, and O. Turan, “A domain-adapted NLP pipeline for human factors classification of maritime accidents,” Ocean Eng., vol. 357, no. P3, p. 125625, 2026, doi: 10.1016/j.oceaneng.2026.125625.
P. Guleria, “NLP based text classification using TF-IDF enabled fine-tuned long short-term memory : An empirical analysis,” Array, vol. 27, no. July, p. 100467, 2025, doi: 10.1016/j.array.2025.100467.
T. Meyer, T. Schlosser, S. Moniri, M. Friedrich, D. Kowerko, and A. Sampath, “Engineering Applications of Artificial Intelligence Medical text generation and classification using large language models : A feasibility study in ophthalmology,” Eng. Appl. Artif. Intell., vol. 178, no. P2, p. 114962, 2026, doi: 10.1016/j.engappai.2026.114962.
Z. M. Koli, J. Alam, Z. Sultana, and A. A. Khan, “BanglaVerb : A sentence-level dataset for transitivity classification in Bangla NLP,” Data Br., vol. 65, p. 112605, 2026, doi: 10.1016/j.dib.2026.112605.
M. Arslan and C. Cruz, “^ ĐŝĞŶĐĞ ŝƌĞĐƚ ScienceDirect Leveraging NLP approaches to define and implement text relevance Leveraging NLP approaches to define and implement text relevance hierarchy framework for business news classification hierarchy framework for business news classification,” Procedia Comput. Sci., vol. 225, pp. 317–326, 2023, doi: 10.1016/j.procs.2023.10.016.
X. Ren, M. Jiang, and Y. Wang, “ScienceDirect Text Style Recognition and Classification Model with Multi- Feature Fusion,” Procedia Comput. Sci., vol. 285, pp. 366–374, 2026, doi: 10.1016/j.procs.2026.07.205.
M. Bourdin, A. Neumann, T. Paviot, R. Pellerin, and S. Lamouri, “A Framework for Selecting the Optimal NLP Solution for Classification Tasks in Industry 4 . 0 Based on Data and Business Constraints,” IFAC Pap., vol. 59, no. 10, pp. 1850–1855, 2025, doi: 10.1016/j.ifacol.2025.09.311.
I. Didier and R. Luiz, “Technovation AI-powered innovation tournament evaluations : A hybrid NLP approach,” vol. 156, no. February, 2026, doi: 10.1016/j.technovation.2026.103622.
V. Kwaku, W. Chen, S. Banaamwini, C. C. Ukwuoma, Y. Hyeon, and M. A. Al-antari, “Heliyon MuTCELM : An optimal multi-TextCNN-based ensemble learning for text classification,” Heliyon, vol. 10, no. 19, p. e38515, 2024, doi: 10.1016/j.heliyon.2024.e38515.
C. A. Mehdi, M. Oubrahim, N. Joudar, and M. Ettaouil, “GEODENSECONVNET : A BREGMAN DIVERGENCE-INSPIRED FRAMEWORK FOR ADAPTIVE FEATURE FUSION IN TEXT CLASSIFICATION,” vol. 38, no. 11, pp. 1908–1919, 2025.
P. He et al., “D E BERT A : D ECODING - ENHANCED BERT ENTANGLED A TTENTION WITH D IS -,” pp. 1–21, 2021.
S. Mallat and M. Zrigui, “ScienceDirect The optimization of deep learning classification through the The optimization of deep learning classification through the comparison between di ff erent algorithm by using LABR comparison between di ff erent algorithm by using,” Procedia Comput. Sci., vol. 246, pp. 3575–3584, 2024, doi: 10.1016/j.procs.2024.09.200.
X. Zhang, J. Zhao, and Y. Lecun, “Character-level Convolutional Networks for Text,” pp. 1–9, 2015.
H. Sebbaq and N. El Faddouli, “MTBERT-Attention : An Explainable BERT Model based on Multi-Task Learning for Cognitive Text Classification,” vol. 21, 2023, doi: 10.1016/j.sciaf.2023.e01799.
J. F. C. M, A. Rizzardi, S. Sicari, and A. C. Porisini, “Computers & Security NERO : NEural algorithmic reasoning for zeRO-day attack detection in the IoT : A hybrid approach,” Comput. Secur., vol. 142, no. April, p. 103898, 2024, doi: 10.1016/j.cose.2024.103898.
E. E. Abdallah, W. Eleisah, A. F. Otoom, and I. I. Ahmed, “ScienceDirect ScienceDirect Intrusion Detection Systems using Supervised Machine Learning Techniques : survey Machine Learning Intrusion Detection Systems using A Supervised Techniques : A survey,” Procedia Comput. Sci., vol. 201, pp. 205–212, 2022, doi: 10.1016/j.procs.2022.03.029.
A. Zohourian, S. Dadkhah, H. Molyneaux, E. Carlos, P. Neto, and A. A. Ghorbani, “Computers & Security IoT-PRIDS : Leveraging packet representations for intrusion detection in IoT networks,” Comput. Secur., vol. 146, no. August, p. 104034, 2024, doi: 10.1016/j.cose.2024.104034.
Z. Wang, Y. Liu, D. He, and S. Chan, “Intrusion detection methods based on integrated deep learning model,” vol. 103, 2021, doi: 10.1016/j.cose.2021.102177.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Ali Jaber Tayh Albderi, Rasha Falah Kadhem
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
