CROSS-LINGUAL SENTIMENT ANALYTICS FOR ELECTRIC VEHICLE FEEDBACK USING TRANSFORMER-BASED SEMANTIC ALIGNMENT

Abstract

The rapid global adoption of electric vehicles (EVs) has generated large volumes of user-generated feedback across online platforms in multiple languages. Understanding public perception from these multilingual reviews is crucial for manufacturers, policymakers, and researchers seeking to improve EV technology and user satisfaction. However, traditional sentiment analysis systems primarily focus on monolingual datasets and often rely on machine translation techniques that distort semantic meaning and sentiment polarity across languages. This creates a significant gap in accurately analyzing multilingual EV feedback. To address this limitation, this study proposes a Transformer-Based Semantic Alignment Framework for cross-lingual sentiment analytics. The framework integrates multilingual transformer embeddings, semantic alignment layers, and EV-specific sentiment feature extraction to unify sentiment representations across languages. Experiments were conducted using multilingual EV review datasets collected from Kaggle and the Multilingual Amazon Reviews Corpus. The proposed model was evaluated against baseline models including Support Vector Machine, CNN, LSTM, and multilingual BERT. Experimental results demonstrate that the proposed approach achieves an accuracy of 96.8%, outperforming existing models by an average improvement of 4–8%. The findings confirm that semantic alignment using transformer architectures significantly enhances cross-lingual sentiment understanding. The proposed framework provides a scalable solution for multilingual opinion mining and offers valuable insights for EV market analysis and consumer perception monitoring.