leekwoon/ai-conferences-paper-recommender_backup

AI Conferences Paper Recommender Backup

A semantic search and recommendation system for papers from major AI conferences, powered by vector embeddings and ChromaDB.

Download & Extract

# huggingface_hub 설치 (필요시)
pip install huggingface_hub

# 다운로드
huggingface-cli download leekwoon/ai-conferences-paper-recommender_backup --repo-type dataset --local-dir ./paper_recommender_data

# 무결성 확인 (선택사항)
cd paper_recommender_data
md5sum -c checksums.md5

# 파일 합치기 및 압축 해제
cat data.tar.gz.part_* | tar -xzvf -

Directory Structure

ai-conferences-paper-recommender/
├── data/                   # Conference paper data (JSON format)
│   ├── CVPR/              # Computer Vision and Pattern Recognition
│   │   ├── CVPR_2021.json
│   │   ├── CVPR_2022.json
│   │   ├── CVPR_2023.json
│   │   ├── CVPR_2024.json
│   │   └── CVPR_2025.json
│   ├── ECCV/              # European Conference on Computer Vision
│   │   ├── ECCV_2022.json
│   │   └── ECCV_2024.json
│   ├── ICCV/              # International Conference on Computer Vision
│   │   ├── ICCV_2021.json
│   │   ├── ICCV_2023.json
│   │   └── ICCV_2025.json
│   ├── ICLR/              # International Conference on Learning Representations
│   │   ├── ICLR_2021.json
│   │   ├── ICLR_2022.json
│   │   ├── ICLR_2023.json
│   │   ├── ICLR_2024.json
│   │   └── ICLR_2025.json
│   ├── ICML/              # International Conference on Machine Learning
│   │   ├── ICML_2021.json
│   │   ├── ICML_2022.json
│   │   ├── ICML_2023.json
│   │   ├── ICML_2024.json
│   │   └── ICML_2025.json
│   └── NeurIPS/           # Neural Information Processing Systems
│       ├── NeurIPS_2021.json
│       ├── NeurIPS_2022.json
│       ├── NeurIPS_2023.json
│       ├── NeurIPS_2024.json
│       └── NeurIPS_2025.json
├── chroma_dir/            # ChromaDB vector database
│   ├── chroma.sqlite3     # SQLite database file
│   └── */                 # Vector index files
├── backup/                # Backup scripts and crawlers
│   ├── paper_spider/      # Scrapy spider for crawling papers
│   └── scripts/           # Crawling scripts for each conference
├── app.py                 # FastAPI application
├── make_chroma.py         # Script to build ChromaDB from JSON data
├── constants.py           # Configuration constants
├── requirements.txt       # Python dependencies
├── Dockerfile            # Docker configuration
└── start.sh              # Application startup script

Key Features

Supported Conferences

• ICML: International Conference on Machine Learning
• NeurIPS: Neural Information Processing Systems
• ICLR: International Conference on Learning Representations
• CVPR: Computer Vision and Pattern Recognition
• ECCV: European Conference on Computer Vision
• ICCV: International Conference on Computer Vision

Data Coverage

• Years: 2021-2025 (varies by conference)
• Total Papers: Thousands of papers with titles, authors, and abstracts
• Update Frequency: Includes latest conference proceedings

API Endpoints

/search-papers (POST)

Search for similar papers based on semantic query.

Parameters:

• query: Search query text
• conferences: List of conferences to search (default: all)
• years: List of years to search (default: 2022-2025)
• recall_top_k: Number of results to return (default: 10)

Response: List of relevant papers with similarity scores

Technologies Used

• FastAPI: Web framework for building APIs
• ChromaDB: Vector database for semantic search
• LangChain: Framework for LLM applications
• OpenAI Embeddings: Text embedding model for semantic similarity
• Scrapy: Web crawling framework for data collection

Setup and Usage

Requirements

• Python 3.8+
• OpenAI API key (for embeddings)
• Docker (optional)

Installation

# Install dependencies
pip install -r requirements.txt

# Set OpenAI API key
export OPENAI_API_KEY="your-api-key"

# Build ChromaDB from JSON data
python make_chroma.py

# Start the application
./start.sh

Docker Deployment

# Build Docker image
docker build -t paper-recommender .

# Run container
docker run -p 8000:8000 -e OPENAI_API_KEY="your-api-key" paper-recommender

Data Format

Each paper JSON entry contains:

{
  "title": "Paper Title",
  "authors": ["Author 1", "Author 2"],
  "abstract": "Paper abstract text...",
  "pdf_link": "URL to PDF",
  "conference": "ICML",
  "year": 2024
}

Use Cases

• Find similar papers to your research
• Discover relevant work across multiple conferences
• Track research trends in specific areas
• Build citation networks
• Generate reading lists for new research areas

Citation

This tool aggregates publicly available conference papers for research purposes. Please cite the original papers and conferences when using this data.