SmartResume 0.1.1

Intelligent Resume Parsing System

SmartResume - Intelligent Resume Parsing System

💻 Code   |   🤗 Model   |   🤖 Demo   |   📑 Technical Report

English | 中文

Project Introduction

SmartResume is an layout‑aware resume parsing system. It ingests resumes in PDF, image and common Office formats, extracts clean text (OCR + PDF metadata), reconstructs reading order with layout detection, and leverages LLMs to convert content into structured fields such as basic info, education, and work experience.

demo

Quick Start

Requirements

• Python >= 3.9
• CUDA >= 11.0 (optional, for GPU acceleration)
• Memory >= 8GB
• Storage >= 10GB

Installation

1. Clone the repository

git clone https://github.com/alibaba/SmartResume.git
cd SmartResume

2. Create conda environment

conda create -n resume_parsing python=3.9
conda activate resume_parsing

3. Install dependencies

pip install -e .

4. Configure environment

# Copy configuration template
cp configs/config.yaml.example configs/config.yaml
# Edit configuration file and add API keys
vim configs/config.yaml

Basic Usage

Method 1: Command Line Interface (Recommended)

# Parse single resume file
python scripts/start.py --file resume.pdf

# Specify extraction types
python scripts/start.py --file resume.pdf --extract_types basic_info work_experience education

Method 2: Python API

from smartresume import ResumeAnalyzer

# Initialize analyzer
analyzer = ResumeAnalyzer(init_ocr=True, init_llm=True)

# Parse resume
result = analyzer.pipeline(
    cv_path="resume.pdf",
    resume_id="resume_001",
    extract_types=["basic_info", "work_experience", "education"]
)

print(result)

Local Model Deployment

SmartResume now supports local model deployment using vLLM, reducing dependency on external APIs:

# Download Qwen-0.6B-resume model
python scripts/download_models.py

# Deploy model
bash scripts/start_vllm.sh

For detailed local model deployment guide, see LOCAL_MODELS.

Key Features

Metric Category	Specific Metric	Value	Description
Layout Detection	mAP@0.5	92.1%	High layout detection accuracy
Information Extraction	Overall Accuracy	93.1%	High accuracy
Processing Speed	Single Page Time	1.22s	High performance
Language Support	Supported Languages	many	Covering major global languages

Benchmark Results

For detailed benchmark results, see Benchmark Results.

Configuration

For detailed configuration options, see the Configuration Guide.

License Information

This project is licensed under the LICENSE.

Currently, some models in this project were previously trained with third-party detectors. We plan to explore and replace them with models under more permissive licenses to enhance user-friendliness and flexibility.

Acknowledgments

• PDFplumber
• EasyOCR

Citation

@article{Zhu2025SmartResume,
  title={Layout-Aware Parsing Meets Efficient LLMs: A Unified, Scalable Framework for Resume Information Extraction and Evaluation},
  author={Fanwei Zhu and Jinke Yu and Zulong Chen and Ying Zhou and Junhao Ji and Zhibo Yang and Yuxue Zhang and Haoyuan Hu and Zhenghao Liu},
  journal={arXiv preprint arXiv:2510.09722},
  year={2025},
  url={https://arxiv.org/abs/2510.09722}
}

---

Note: Please ensure compliance with relevant laws and regulations and privacy policies.