AI-powered system for detecting fake reviews, analyzing sentiment, and generating summaries. Combines BERT, embeddings, and deep learning to ensure trustworthy insights for e-commerce and service domains.
This is an AI-driven system that detects fake reviews, performs sentiment analysis, and generates concise summaries to support trustworthy decision-making. Built with a hybrid model of BERT, character embeddings, and token embeddings, the system is designed for use in e-commerce and service-oriented platforms where user feedback heavily influences consumer behavior.
Features:
- Detects fake reviews using deep learning and linguistic patterns
- Hybrid NLP model using BERT + token/character embeddings
- Sentiment analysis (polarity, subjectivity, etc.)
- Review summarization with keyword/polarity filtering
- Review screening with personalization options
- Outputs real, trustworthy reviews with summaries
- Easy CSV-based input/output for batch processing
Tech Stack:
- Python
- TensorFlow / Keras
- BERT (via Hugging Face Transformers)
- NLTK / TextBlob
- Flask (for deployment)
- Pandas / NumPy
- HTML/CSS/Js
NOTE:
-
This project uses DeepSeek7B (https://ollama.com/library/deepseek-llm) via Ollama (https://ollama.com/) for text summarization. Make sure you have Ollama installed and the DeepSeek model pulled.
-
This project uses Oxylabs' Web Scraper API for scraping amazon reviews. Make sure you have the required details in order to use the API.
-
Some parts of the code are placeholders where you need to insert your own values, such as:
- API keys (Oxylabs)
- Usernames and passwords
- IP addresses or hostnames (e.g., for local servers like Ollama)
Make sure to fill in these values before running the project.