Skip to content

Web application for summarizing and asking questions on your own documents. Summarization and QA application in FastAPI, React and Transformers

License

Notifications You must be signed in to change notification settings

MarkusSagen/Surmize

Repository files navigation

Surmize

A full scale client-server model for Summarizing and asking close-domain Questions on documents and or texts directly in your browser. The file extentions .txt, .story and .pdf are currently supported and some .CSV-files.

This project was developed by us for the course Independent Project in Information Engineering at Uppsala University. The applicaiton is built mostly in React, FastAPI and Python modules. The underlying NLP models are based on Huggingface. Both models are state-of-the-art transformer models, which have performed very well on a diverse number of NLP tasks.

Landing Page

Work Area

Usage

These instructions are related to how to get the server up an running, the UI and NLP models for summarization and Close-Domain QA.

Some have experienced problems when using Python3 directly, we therefore recommend users to install and use Python3 via Anaconda

  1. Clone this repo:
git clone https://github.com/MarkusSagen/Surmize.git
  1. Make sure Java OpenJDK is installed.
  2. Run the setup script:
source start.sh
  1. Start up the server in development mode by running:
uvicorn api:app --reload --port 5000
  1. Open another terminal and navigate to the folder Surmize/client/ and run:
npm install && npm start
  1. Go to the address http://localhost:3000 in a browser

  2. Upload files or texts you want to ask questions about, example texts can be found here

Citation

@misc{Bergkvist1436450,
   author = {Bergkvist, Alexander and Hedberg, Nils and Rollino, Sebastian and Sagen, Markus},
   institution = {Uppsala University, Department of Information Technology},
   pages = {56},
   school = {Uppsala University, Department of Information Technology},
   title = {Surmize: An Online NLP System for Close-Domain Question-Answering and Summarization},
   series = {Självständigt arbete i informationsteknologi},
   number = {2020-001},
   year = {2020}
}

Contributors

Markus Sagen
Alexander Bergkvist
Nils Hedberg
Sebastian Rollino

About

Web application for summarizing and asking questions on your own documents. Summarization and QA application in FastAPI, React and Transformers

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published