metadata
pretty_name: Cue Benchmark v0.1
license: other
license_name: mixed-license
license_link: >-
https://huggingface.co/datasets/SporeLabs/cue-benchmark/blob/main/LICENSE-NOTES.md
task_categories:
- text-classification
language:
- en
tags:
- benchmark
- meetings
Cue benchmark v0.1
Test items, scorer and results for cues in meetings: commitments and open questions, scored against human labels from the AMI and ICSI meeting corpora. No audio is included.
Start with the benchmark card for task definitions, gold provenance, sample sizes, operating points and baselines. Read the licence notes before reusing any item.
hf download SporeLabs/cue-benchmark --repo-type dataset --local-dir cue-benchmark
cd cue-benchmark
python scorer/run_score.py --preds preds_example --out out
See the scoring instructions and results. The model released with this benchmark is cue-text.