Home

textsum wiki

Welcome to the textsum wiki!

textsum is a utility package that allows you to apply transformer-based summarization models to text documents of arbitrary length. The package includes a command line interface (CLI) and a user interface demo using gradio.

It was originally built as an extension/generalization of the document summarization space on huggingface.

Contents

The purpose of this wiki is to document how to practically use transformer models for summarization, what the parameters do, and how to improve performance.

1. inference parameters - what do they do, what values to use.
2. Examples and demos - A (small) list of demos and examples.
3. consolidating summaries - how to use a secondary text2text model for map-reduce summarization
4. synthsumm - summarization models trained on synthetic data show improved generalization

Installation

The package can be installed using pip from pypi:

pip install textsum

Please have a look at the README for more details/options.