Overview
ResearchRec is a platform that provides personalized research article recommendations by integrating advanced machine learning techniques like Active Learning and Transfer Learning. It ensures dynamic adaptation to user preferences and effective handling of cold-start scenarios.
Key Features
Trending Articles
View globally popular and engaging articles in a dynamic list.
Personalized Recommendations
Tailored research article suggestions based on user preferences and interactions.
Interactive User Feedback
Refine recommendations dynamically through user feedback and interactions.
Real-Time Updates
Receive live updates on trending articles and personalized recommendations.
Advanced Machine Learning Integration
Incorporates NCF and Active Learning for enhanced recommendation accuracy.
Technical Architecture
This platform addresses the growing challenge of navigating the vast volume of scientific literature. It integrates advanced scrapping tools for metadata analysis, user behavior modeling, and generating research article recommendations.
Implementation
The implementation begins with data preprocessing using techniques like TF-IDF vectorization and embeddings for extracting features from abstracts and tags.
Neural Collaborative Filtering (NCF) is employed to model complex interactions between users and articles. The Active Learning module enhances adaptability by iteratively refining the model with high-uncertainty data points, reducing dependency on large labeled datasets.
Tech Stack
- Frontend: Next.js provides a responsive interface
- Backend: Nest.js powers efficient server-side processing, handles REST APIs, manages the database
- AI Service: Flask manages the recommendation model and facilitates integration
- Database: MongoDB stores user data, interactions, and metadata
- ML Framework: TensorFlow for deep learning models
Features
A hybrid scoring mechanism balances personalized recommendations with global article popularity, ensuring relevance and diversity. The application enables real-time updates on trends and incorporates user feedback loops for continuous improvement.