Project Title: Flipkart Recommendation Using Mobile Reviews with LLM and NLP Goal: Developed a personalized product recommendation system by leveraging Large Language Models (LLMs) and Natural Language Processing (NLP) to analyze and generate insights from customer reviews on Flipkart.
Process:
- Data Collection:
Scraped customer reviews and product details from Flipkart using web scraping techniques (e.g., Selenium, BeautifulSoup). Gathered datasets containing customer sentiments, review texts, product features, and ratings.
- Data Preprocessing:
Applied NLP preprocessing techniques such as tokenization, stopword removal, and stemming/lemmatization. Used BERT embeddings to convert textual data into numerical vectors for advanced analysis.
3 .Sentiment Analysis:
Built a sentiment analysis model leveraging LLMs like OpenAI’s GPT and Hugging Face Transformers for robust sentiment classification. Categorized reviews as Positive, Neutral, or Negative.
- Recommendation System Development:
Trained a collaborative filtering model augmented by NLP insights from review sentiment. Used the Random Forest Regressor and fine-tuned LLMs to predict and recommend products tailored to user preferences.
- Visualization and Deployment:
Designed an intuitive Streamlit application to display recommendations dynamically. Included interactive filters and visualization tools (e.g., word clouds, sentiment trends). Deployed the project on GitHub and AWS for accessibility and scalability.
Outcome: Delivered a robust end-to-end product recommendation system that improves user experience by analyzing mobile reviews and predicting preferences with precision. Achieved enhanced accuracy and customer satisfaction using LLM-driven insights. The project is done on GitHub for collaboration and future iterations.