Plat form Windows 11 Anaconda
import nltk as nltk
nltk.__version__
'3.8.1'
The sentence_bleu ought to return 1 for identical translation
from nltk.translate.bleu_score import sentence_bleu, SmoothingFunction
sentence_bleu([["hi", "hello", "world"]], ["hi", "hello", "world"])
1.2213386697554703e-77
and even with a smooth function it could not help much
smoother = SmoothingFunction()
sentence_bleu([["hi", "hello", "world"]], ["hi", "hello", "world"], smoothing_function=smoother.method4)
0.5757197301274735
However,
sentence_bleu([["hi", "hello", "world", "how"]], ["hi", "hello", "world", "how"])
1.0
This appeared to be a bug in case handling or the summation index.