Review Lens

Problem Statement

Design and implement an ML-based system to evaluate the quality and relevancy of Google location reviews.

Project Overview

Our ML project, Review Lens is an automated tool that classifies location-based reviews related to Food and Beverages into 4 categories namely, "Advertisement","Spam","Valid Review" and "Not Related to F&B". This solution addresses the challenge of assessing review quality by streamlining the analysis of large datasets. It provides business owners with clear, actionable insights by filtering out irrelevant reviews and highlighting valuable feedback.

Key Features:

Classification: It automatically classifies reviews into specific categories like "Valid Review," "Spam," "Advertisement," or "Not Related to F&B." This provides immediate insight into the nature of the feedback, helping businesses quickly sort through and prioritize reviews. Relevancy Scoring: For reviews identified as valid, the system calculates a numerical relevancy score. This score quantifies how much a review's content aligns with the business's categories and attributes, helping businesses pinpoint feedback on key aspects of their service. Spam Filtering: Flags a review if the same reviewer has repeatedly reviewed the same business more than 5 times

Tools Used

Development tools used: Colab, VScode APIs used: Qwen Libraries and frameworks used: Hugging Face Transformers, pandas, duckdb

Assets and datasets used:

• Yelp businesses,reviews and users dataset from https://www.kaggle.com/datasets/yelp-dataset/yelp-dataset
• Scraped Top 10 restaurant google reviews for California and Florida

Built With

• colab
• duckdb
• huggingface
• pandas
• qwen
• vscode