summary-workbench
summary-workbench copied to clipboard
Framework for unified summarisation and evaluation of English documents using state-of-the-art models and measures.
Summary Workbench
Unifying the application and evaluation of text summarization models. [Paper] [Documentation] [Live Demo]
Accepted at EMNLP 2022 (Demo track). :tada: :tada:
:loudspeaker: Updates (3-12-2022)
- Integrated 2 new models and their variants (6 in total): BRIO trained on news, Schnitsum trained on scholary documents.
- Integrated contrastive search for more fluent summaries. User can now toggle between regular and contrasitve search for supported models.
- Improvements to the UI responsiveness on smaller devices.
Summarize
Create a Request

Inspect the Results

Evaluate
Create a Request

Inspect the Results
Scores

Visualize Text Examples

Plot Scores against each other
