Skip to content

Scientific article summarization

Note

GitHub Model Dataset

Technical skill used

  • Python
  • Tensorflow
  • Transformers
  • nltk
  • HuggingFace

Synopsis

This model generate a summarized scientific article from arXiv. The goal is to understand in a few sentences the content of the article. It is not aimed to replace the article abstract.

Model Description

The model used is a T5-small with Hugging-Face weight (Transfert Learning)

Inputs

  • text as a token format

outputs

  • A text with a number of word. The general meaning is as close as the article abstract.

Model limitation

  • The article must be available at arXiv.
  • The article must come in tex format.
  • The `tex' should be in a valid format.
  • The article should be written in English.

Tradeoff

  • A complete resume requires a longer inference time.
  • Mathematics formulas and citations are hidden to not have to make another train for understanding these formulas