/arxiv-daily-summaries

An issue a day keeps the paper overload away.

Primary LanguagePython

arXiv Daily Summaries

This project creates an issue on every work day with the latest arXiv pre-prints matching a given set of keywords. It also creates a TL;DR summary automatically using BART by default.

Configuration

In config.json file, you can specify:

  • arxiv_base: The base URL to arXiv.org. Default to "https://arxiv.org".
  • sub_url: Where to look for paper. Default to "https://arxiv.org/list/new/cs".
  • enable_emojis: Whether to enable emojis or not. Default to False.
  • keywords: List of keywords to look for in each papers. Those selected will get a generated summaries from the model in model_name.
  • assignees: List of assignees whenever a new issue is created. Default to os.environ['GITHUB_REPOSITORY_OWNER'].
  • tldr_max_length: Max length for the generated summaries.
  • model_name: Model name for generating text.

Usage

1. Fork it to your repository on Github

2. Update config.json

{
  "enable_emojis": true,
  "keywords": ["iot"],
  "assignees": ["danielphan2003"],
  "tldr_max_length": 200
}

3. Run a orkflow

To test the functionality, you can click "Run Workflow button" for an immediate run.