Integrating Technical Analysis into Sentiment Analysis: An ASTE Framework for Electric Car Purchase Decision Support Based on LLMs and Semantic BNF

The increasing complexity of purchasing an electric car, influenced by technical specifications and expert reviews, requires advanced Natural Language Processing techniques to extract meaningful insights. This study enhances Aspect Sentiment Triplet Extraction (ASTE) by integrating Large Language Models (LLMs) to identify key aspects, opinions, and sentiments in expert reviews, including technical data traditionally classified as neutral, such as horsepower and battery range. A semantic extension of Backus-Naur Form (BNF) structures input queries for syntactic and semantic accuracy, while a 2-tuple fuzzy linguistic model refines sentiment representation, ensuring interpretability. The proposed model addresses limitations in existing ASTE techniques by incorporating formal grammar structures and linguistic modeling, eliminating the need for complex preprocessing. Applied to expert YouTube reviews of electric cars, the method leverages Google's Gemini model via Python and the Gemini API to rank the top-selling electric cars in the United States. The results confirm the model's effectiveness in aligning technical data with sentiment analysis, making it accessible to non-specialists in Natural Language Processing. This framework enhances decision support in electric car purchases by providing a structured, interpretable, and contextually rich sentiment analysis approach.