Skip to content

Latest commit

 

History

13 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Div-Ref

The official repository of "Not All Metrics Are Guilty: Improving NLG benchmarks by Diversifying References".

The references and diversified references can be found in dataset/XXX/xxxx.reference and dataset/XXX/xxxx.reference.diversified.

Requirements

git clone https://github.com/google-research/mt-metrics-eval.git
cd mt-metrics-eval
pip install .
cd ..

git clone https://github.com/google-research/bleurt.git
cd bleurt
pip install .
cd ..

git clone https://github.com/thompsonb/prism
cd prism
pip install -r requirements.txt
cd ..

pip install -U git+https://github.com/pltrdy/pyrouge
git clone https://github.com/pltrdy/files2rouge.git     
cd files2rouge
python setup_rouge.py
python setup.py install
cd ..

git clone https://github.com/neulab/BARTScore
git clone https://github.com/AIPHES/emnlp19-moverscore

pip install sacrebleu, bert-score, unbabel-comet, pycocoevalcap

WMT22 Metrics Shared Task

Download data:

alias mtme='python3 -m mt_metrics_eval.mtme'
mtme --download

Prepare data:

python data.py zh-en

Calculate metric scores for different metrics:

python score.py score/zh-en.json chrf/bleu/bertscore/bleurt/prism/comet/bartscore/gemba

Calculate the correlation scores for different metrics:

python analysis.py score/zh-en.json chrf/bleu/bertscore/bleurt/prism/comet/bartscore/gemba

SummEval

Download data:

wget https://storage.googleapis.com/sfr-summarization-repo-research/model_annotations.aligned.jsonl

Prepare data:

python data.py

Calculate metric scores for different metrics:

python score.py score/summeval.json rouge/bertscore/moverscore/chatgpt

Calculate the correlation scores for different metrics:

python analysis.py score/summeval.json rouge/bertscore/moverscore/chatgpt

PASCAL-50S

Download data:

wget https://filebox.ece.vt.edu/~vrama91/CIDEr_miscellanous/cider_datasets.tar

Prepare data:

python data.py

Calculate metric scores for different metrics:

python score.py score/pascal.json bleu/rouge/meteor/cider/spice/bertscore

Calculate the correlation scores for different metrics:

python analysis.py score/pascal.json bleu/rouge/meteor/cider/spice/bertscore

About

The official repository of "Not All Metrics Are Guilty: Improving NLG Evaluation Diversifying References".

Resources

Stars

3 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages