Part 2/9:
The process begins with sourcing large datasets of product reviews, primarily from platforms like Kaggle and Google Dataset Search. One notable dataset targeted is the "Amazon Product Reviews," which contains over 34,000 reviews spanning diverse product categories.
The reviews come as extensive CSV files, with many fields such as URLs, ratings, and textual reviews. For meaningful analysis, the focus narrows to the core components: the product name, review title, and the review text itself. Cleaning involves filtering out irrelevant reviews (e.g., charger-specific reviews when analyzing a device) and converting text encoding issues to ensure compatibility with Python scripts.