<p>Customer reviews available on e-commerce websites and social media play a significant role on purchasing decisions and improving product quality. However, many existing sentiment analysis methods often face difficulties in accurately identify capture aspect-based sentiments and handle mixed or unclear opinions. To address these issues, this paper proposes a novel sentiment classification framework FLS-FFA that combines a Fuzzy Logic System (FLS) with the Fennec Fox Algorithm (FFA). The review data collected from the IMDB, Stanford Sentiment Treebank, Yelp Polarity, and Amazon Prime Movies datasets is preprocessed through comprehensive pipeline to remove noise and improve data quality. A hybrid feature extraction method combines review-level and aspect-level information, while the Puma Optimization algorithm (POA) selects the most suitable features to reduce classification complexity. Subsequently, a Bi-Path Text-to-Text Transfer Transformer Network (BPT5N) generates sentiment scores, which are analyzed using a fuzzy logic inference mechanism for sentiment classification. The FFA is applied to optimize fuzzy membership functions and reduce ambiguity among sentiment classes. Experimental results demonstrate the effectiveness of the proposed framework, achieving classification accuracies of 96.30%, 97.00%, 96.80%, and 96.01% on the IMDB, Stanford Sentiment Treebank, Yelp Polarity, and Amazon Prime Movies datasets, respectively. These findings outperform existing state-of-the-art methods and confirm the robustness, reliability, and practical applicability of the proposed FLS-FFA framework for sentiment analysis and customer feedback interpretation.</p>