E-commerce Multi-Source Product Review Analysis & QA Bot

🧾 Overview

This project tackles the challenge of understanding customer feedback from diverse sources — including Amazon, YouTube, and Reddit — by aggregating, clustering, and analyzing product reviews across platforms.

In addition to review aggregation and topic modeling, this system also implements a QA bot that allows users to ask product-specific questions and receive context-driven answers sourced from actual user reviews using a Retrieval-Augmented Generation (RAG) architecture.

The result is an end-to-end product insight system designed to help:

• Consumers make better purchase decisions
• Product managers detect sentiment patterns
• Analysts discover trending product themes

🚧 Problem Statement

Product reviews are scattered across platforms and filled with noise, redundancy, and unstructured sentiment. While individual reviews are helpful, they lack aggregated insight or structured Q&A capabilities.

Challenges addressed:

• 🚫 Inconsistent review formats across platforms (text, video, Reddit posts)
• 🧩 Difficulty finding thematic consensus across thousands of reviews
• ❓ Inability to ask contextual questions like:
  “What do users say about battery life?” or “Is the camera good in low light?”

This project solves those gaps using:

• Multi-source scraping
• Topic modeling and clustering
• QA via semantic search + LLM-generated answers

🧠 Project Pipeline

🛒 Data Collection:

• YouTube transcripts scraped via youtube_transcript_api
• Amazon reviews (pre-collected dataset)
• Reddit product threads using PRAW

🔍 Review Preprocessing:

• Regex cleaning and normalization
• Sentence filtering
• Deduplication and source labeling

🧾 QA Bot (RAG):

• FAISS vector store for document retrieval
• LangChain/OpenAI for contextual QA generation
• Top-k similarity-based passage ranking

Result: Ask “What do people dislike about this product?” and get a grounded, review-based answer.

🧪 Example QA Interactions

Query: “What do users say about battery life of iPhone 15?”
Answer:

• Most users report that the battery lasts about 6–8 hours of heavy use.
• Several Reddit posts highlight fast battery drain after 1 year.
• YouTube reviewers suggest turning off background sync to save power.

---

Query: “Is the camera good in low light?”
Answer:

• Amazon reviews frequently mention poor detail in night shots.
• Reddit users recommend using manual mode for better results.
• A YouTube reviewer compares it unfavorably with a mid-tier DSLR.

🎯 Key Features

• 🔄 Multi-Source Scraping: Review data from Reddit, YouTube, and Amazon
• 🧼 Smart Preprocessing: Text cleaning, deduplication, and sentence-level segmentation
• 🧠 Topic Modeling with BERTopic: Identify pain points and praise patterns
• 🤖 QA Bot Powered by RAG: Ask questions and get real, evidence-based answers
• ⚡ Modular Design: Each step can be used independently or chained