This paper presents the design and implementation of a machine-learning-based customer review analysis and sum-marization system optimized for local businesses in Hyderabad. Unlike traditional review platforms that overwhelm users with raw textual feedback, the proposed architecture adopts an intelligent analytical pipeline in which critical tasks—automatic review collection, sentiment classification, and aspect-based sum-marization—are executed to generate concise and meaningful business insights. The system integrates transformer-based sen-timent classification (BERT/DistilBERT), sequence-to-sequence abstractive summarization (BART), extractive summarization (TextRank), TF-IDF keyword extraction, and spaCy named entity recognition (NER) within a unified, modular pipeline deployed through a Streamlit web application. A structured summarization framework extracts key business strengths and weaknesses, overall customer sentiment, and focused feedback through an intuitive, user-friendly interface. Experimental eval-uation on a custom dataset of 12,500 Hyderabad local business reviews demonstrates a macro-averaged F1 score of 0.881 and overall accuracy of 89.3% for sentiment classification, alongside a BART summarization ROUGE-1 score of 0.456. The system demonstrates the potential of lightweight, scalable, and intelligent review analysis for transparent, reliable, and data-driven business decision support across diverse local business use cases.

Index Terms—sentiment analysis, text summarization, BERT, DistilBERT, BART, TF-IDF, natural language processing, local business intelligence, Hyderabad reviews, transformer models