A modern, intelligent machine learning paper recommendation system that helps researchers discover relevant papers tailored to their interests.
- Intelligent Recommendations: Content-based recommendation algorithm using TF-IDF and cosine similarity to suggest papers based on your preferences
- Hierarchical Subject Filters: Filter papers by ML subfields (NLP, Computer Vision, Reinforcement Learning, etc.)
- Boolean Search: Advanced search with AND, OR, and NOT operators for precise queries
- Paper-Type Filters: Filter by Conference, Journal, Preprint, or Workshop papers
- Key Points Extraction: Automatically extracts 3-5 key points from paper abstracts
- Code Availability Badges: Identifies papers with available code repositories
- Persistent Filters: Your filter preferences are saved and restored across sessions
- Minimalist Design: Clean, scannable card-based UI with key metrics
- Continuous Feedback: Like/dislike papers to improve recommendations
- Framework: Next.js 14+ (App Router)
- Language: TypeScript
- Styling: Tailwind CSS
- State Management: Zustand
- Data Source: ArXiv API
- Recommendation Algorithm: Content-based using TF-IDF vectorization
- Node.js 18+ and npm
- Clone the repository:
git clone https://github.com/yourusername/PaperTime.git
cd PaperTime- Install dependencies:
npm install- Run the development server:
npm run dev- Open http://localhost:3000 in your browser.
- Get Recommendations: Click the "Recommend Next" button to get personalized paper recommendations
- Search: Type keywords in the search bar and press Enter, or use boolean operators (AND, OR, NOT)
- Filter: Use the sidebar filters to narrow down by subject, paper type, or search query
- Provide Feedback: Like or dislike papers to improve future recommendations
- Explore: Click on paper titles to view them on ArXiv, or download PDFs directly
This project is configured for GitHub Pages deployment using GitHub Actions.
-
Enable GitHub Pages:
- Go to your repository Settings → Pages
- Under "Source", select "GitHub Actions"
-
Repository Name:
- The GitHub Actions workflow automatically detects your repository name
- If your repository is not named "PaperTime", the workflow will still work correctly
- The site will be available at
https://yourusername.github.io/your-repo-name/
-
Deploy:
- Push to the
mainbranch - GitHub Actions will automatically build and deploy your site
- Your site will be available at
https://yourusername.github.io/PaperTime/
- Push to the
To build the static site locally:
npm run buildThe static files will be in the out directory.
PaperTime/
├── app/
│ ├── components/ # React components
│ │ ├── FilterPanel.tsx
│ │ ├── PaperCard.tsx
│ │ ├── RecommendButton.tsx
│ │ └── SearchBar.tsx
│ ├── lib/ # Utility functions
│ │ ├── api-client.ts # Client-side API functions
│ │ ├── arxiv.ts # ArXiv API client
│ │ ├── filters.ts # Filter logic
│ │ ├── recommendation.ts # Recommendation algorithm
│ │ ├── summarization.ts # Key points extraction
│ │ └── tfidf.ts # TF-IDF implementation
│ ├── stores/ # State management
│ │ └── filterStore.ts
│ ├── layout.tsx
│ └── page.tsx # Main page
├── .github/
│ └── workflows/
│ └── deploy.yml # GitHub Actions workflow
├── next.config.js
├── package.json
└── README.md
The application makes API calls to ArXiv from the browser. Since ArXiv's API doesn't support CORS, the application uses a CORS proxy service (api.allorigins.win) to bypass this restriction. This allows the application to work entirely client-side without requiring a backend server.
If you encounter issues with the CORS proxy, you can:
- Set up your own CORS proxy server
- Use a different CORS proxy service
- Deploy a backend API to proxy requests
The application uses browser-native APIs (DOMParser for XML parsing) and is fully client-side, making it compatible with all modern browsers.
- Feature Extraction: Papers are converted to TF-IDF vectors based on their titles and abstracts
- Similarity Calculation: Cosine similarity is computed between papers
- Ranking: Papers are ranked by:
- Similarity to liked papers (50%)
- Recency (30%)
- Foundational importance (20%)
- Subject Filters: Maps hierarchical ML subjects to ArXiv categories
- Paper Type: Infers type from ArXiv comments and metadata
- Boolean Search: Parses AND/OR/NOT queries and matches against paper content
- Splits abstracts into sentences
- Scores sentences by length, keyword density, and position
- Selects top 3-5 sentences as key points
Contributions are welcome! Please feel free to submit a Pull Request.
ISC
- ArXiv for providing the paper database
- Next.js team for the excellent framework
- All the researchers whose papers make this tool useful