Methodology and Experimental Implementation of a Hybrid BERT–Quantum Model for Text Classification
Contributors
Dr. E.Chandra Blessie
Dr.P.Rubini
Keywords
Proceeding
Track
General Track
License
Copyright (c) 2026 Sustainable Global Societies Initiative

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Abstract
Text classification plays a significant part in inevitably establishing and examining huge volumes of textual information. This paper proposes a hybrid quantum-classical text classification framework that merges Bidirectional Encoder Representations from Transformers (BERT), Principal Component Analysis (PCA), Variational Quantum Circuits (VQC), and a Hybrid Attention mechanism for multi-class news classification. The AG News dataset comprises 6,000 balanced news articles over four classes: World, Sports, Business, and Sci/Tech. We use BERT to extract contextual representations of the news text, producing 768-dimensional feature vectors.
PCA reduces these features to eight dimensions to enable efficient processing using an eight-qubit quantum circuit. The reduced features are encoded into quantum states and processed using a VQC with trainable parameters and entangling operations. The proposed hybrid attention mechanism combines and emphasizes informative classical and quantum features before classification. The performance is compared with classical BERT, PCA-based classical, quantum, and VQC-based models. The proposed approach demonstrates the potential of integrating transformer-based language representations with quantum computing and attention mechanisms for effective text classification.