paper_summarization 🦜️🔗

This work for paper summarization. Built on Streamlit.

LLM model:

• Llama 3 🦙 (~9GB VRAM)
• GPT-4o
• GPT-4-turbo
• GPT-3.5-turbo

Install 🔨

source scripts/install.sh

Recommendation: Python 3.10, Rtx 4090, Cuda 12.2

Download LLama3 🦙

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3

Please refer to ChatOllama for more details.

Run

source scripts/run.sh

How

1. Scrape Data

Select the conference and year, and type your subject of interest (e.g., "hand").

Conference: CVPR, WACV, ICCV

Year: 2011~2024

Click the 'Scrape Data' button on the left side. Then, the paper list will be generated.

2. Save to Excel

Click the 'Save to Excel' button to download the dataframe of the paper list as an Excel file.

The columns of the DataFrame are: Title, Authors, PDF link, Supp link, and Bibtex.

3. Save PDFs

Click the 'Save PDFs' button to download the PDF files of the papers in each folder.

4. Save Supps

Click the 'Save Supps' button to download the supplementary data of the papers in each folder.

5. Summary

Click the 'Summary' button to generate a markdown file and a Word file (docx) in each folder, containing the summary content.

Additional

1. If you use OPENAI's LLMs, then please type your API key. OPENAI API key
2. You can add or remove the supplement PDF file for summary.

Save Folder 📁

The Excel, PDF, supplement data and summary files will be created in material/{conference_name}/{year}/{query}/....

Please note that you should regularly empty the .cache folder.

Question Prompts ❓

You can change the question prompts in prompt.py. I borrowed them from Yunseong.