fastchrf / README.md
jvamvas's picture
Deploy fastChrF Gradio demo with default and aggregate modes
3f5326b verified
|
Raw
History Blame Contribute Delete
1.08 kB

A newer version of the Gradio SDK is available: 6.22.0

Upgrade
metadata
title: fastChrF
emoji: 
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.15.1
app_file: app.py
short_description: Fast sentence-level ChrF for MBR decoding
python_version: '3.12'

fastChrF Space

Interactive demo for fastChrF: fast sentence-level ChrF computation for Minimum Bayes Risk decoding.

Modes

  • Default (pairwise): Computes ChrF between each hypothesis and each reference. Output shape: (num_hypotheses, num_references).
  • Aggregate: Computes a streamlined aggregate ChrF score per hypothesis across all references. Output shape: (num_hypotheses,).

Enter one hypothesis or reference per line. Advanced options mirror the Python API (char_order, beta, remove_whitespace, eps_smoothing).

Citation

@misc{vamvas-sennrich-2024-linear,
      title={Linear-time Minimum Bayes Risk Decoding with Reference Aggregation},
      author={Jannis Vamvas and Rico Sennrich},
      year={2024},
      eprint={2402.04251},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}