Quantum-Accelerated XAI Transformer for Telugu Restaurant Recommendation with LoRA-IndicBERT and Quantum Attention
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This work proposes quantum-accelerated XAI in the context of Telugu restaurant recommendation using transformer architecture with the use of quantum attention mechanism. Our novel methodology moulded IndicBERT fine-tuned with LoRA (Low-Rank Adaptation) as the excellent pre-processing model able to address 98.4% of damage caused by dirty text among 25,000 Telugu reviews replacing regular and well utilized LSTM with Q-BERT models -Quantum Transformer – exhibiting an accuracy of 96.7% and F-1 score as high as 0.951. The system realizes (8 heads, 97.2% accuracy) Multi-Head Quantum Attention mechanisms for contextual information, quantum self-attention layers, a quantum-calibrated TF-IDF feature domain transformation based on answer scores (97.89% accuracy, 95.67% coverage), an adaptive OPTICS clustering approach in which the stopping criteria were tuned with whale optimization algorithms (91.23%, 0.934 AUC-ROC). Cross-lingual transfer learning using MuRIL-based multilingual representations with 94.8% of the semantics preserved. Improved components on XAI that visualize attention lead to higher accuracy (from 85.5% to 97.3%), trust from the user's perspective goes from 75.2% to 95.6%, and interpretability scores go from 78.4% to 93.2%. Statistical verification (F-Value: 312.8934, P-Value: 0.00005, RMSE: 0.0967) also established better performance. The quantum-transformer framework proposed in this study preserves the cultural-linguistic background of users and offers transparent high-precision suggestions.
Fields of Study & Subjects
Explainable Artificial Intelligence (XAI)
Recommender Systems and Techniques
Topic Modeling
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 30 March 2026
- Language
- EN