nltk.translate.gleu_score module

GLEU score implementation.

nltk.translate.gleu_score.corpus_gleu(list_of_references, hypotheses, min_len=1, max_len=4)

Calculate a single corpus-level GLEU score (aka. system-level GLEU) for all the hypotheses and their respective references.

Instead of averaging the sentence level GLEU scores (i.e. macro-average precision), Wu et al. (2016) sum up the matching tokens and the max of hypothesis and reference tokens for each sentence, then compute using the aggregate values.

From Mike Schuster (via email):

“For the corpus, we just add up the two statistics n_match and

n_all = max(n_all_output, n_all_target) for all sentences, then calculate gleu_score = n_match / n_all, so it is not just a mean of the sentence gleu scores (in our case, longer sentences count more, which I think makes sense as they are more difficult to translate).”

>>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',
...         'ensures', 'that', 'the', 'military', 'always',
...         'obeys', 'the', 'commands', 'of', 'the', 'party']
>>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',
...          'ensures', 'that', 'the', 'military', 'will', 'forever',
...          'heed', 'Party', 'commands']
>>> ref1b = ['It', 'is', 'the', 'guiding', 'principle', 'which',
...          'guarantees', 'the', 'military', 'forces', 'always',
...          'being', 'under', 'the', 'command', 'of', 'the', 'Party']
>>> ref1c = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',
...          'army', 'always', 'to', 'heed', 'the', 'directions',
...          'of', 'the', 'party']

>>> hyp2 = ['he', 'read', 'the', 'book', 'because', 'he', 'was',
...         'interested', 'in', 'world', 'history']
>>> ref2a = ['he', 'was', 'interested', 'in', 'world', 'history',
...          'because', 'he', 'read', 'the', 'book']

>>> list_of_references = [[ref1a, ref1b, ref1c], [ref2a]]
>>> hypotheses = [hyp1, hyp2]
>>> corpus_gleu(list_of_references, hypotheses) 
0.5673...

The example below show that corpus_gleu() is different from averaging sentence_gleu() for hypotheses

>>> score1 = sentence_gleu([ref1a], hyp1)
>>> score2 = sentence_gleu([ref2a], hyp2)
>>> (score1 + score2) / 2 
0.6144...

Parameters

• list_of_references (list(list(list(str)))) – a list of reference sentences, w.r.t. hypotheses

• hypotheses (list(list(str))) – a list of hypothesis sentences

• min_len (int) – The minimum order of n-gram this function should extract.

• max_len (int) – The maximum order of n-gram this function should extract.

Returns

The corpus-level GLEU score.

Return type

float

nltk.translate.gleu_score.sentence_gleu(references, hypothesis, min_len=1, max_len=4)

Calculates the sentence level GLEU (Google-BLEU) score described in

Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V. Le, Mohammad Norouzi, Wolfgang Macherey, Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, Jeff Klingner, Apurva Shah, Melvin Johnson, Xiaobing Liu, Lukasz Kaiser, Stephan Gouws, Yoshikiyo Kato, Taku Kudo, Hideto Kazawa, Keith Stevens, George Kurian, Nishant Patil, Wei Wang, Cliff Young, Jason Smith, Jason Riesa, Alex Rudnick, Oriol Vinyals, Greg Corrado, Macduff Hughes, Jeffrey Dean. (2016) Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation. eprint arXiv:1609.08144. https://arxiv.org/pdf/1609.08144v2.pdf Retrieved on 27 Oct 2016.

From Wu et al. (2016):

“The BLEU score has some undesirable properties when used for single

sentences, as it was designed to be a corpus measure. We therefore use a slightly different score for our RL experiments which we call the ‘GLEU score’. For the GLEU score, we record all sub-sequences of 1, 2, 3 or 4 tokens in output and target sequence (n-grams). We then compute a recall, which is the ratio of the number of matching n-grams to the number of total n-grams in the target (ground truth) sequence, and a precision, which is the ratio of the number of matching n-grams to the number of total n-grams in the generated output sequence. Then GLEU score is simply the minimum of recall and precision. This GLEU score’s range is always between 0 (no matches) and 1 (all match) and it is symmetrical when switching output and target. According to our experiments, GLEU score correlates quite well with the BLEU metric on a corpus level but does not have its drawbacks for our per sentence reward objective.”

Note: The initial implementation only allowed a single reference, but now

a list of references is required (which is consistent with bleu_score.sentence_bleu()).

The infamous “the the the … ” example

>>> ref = 'the cat is on the mat'.split()
>>> hyp = 'the the the the the the the'.split()
>>> sentence_gleu([ref], hyp)  
0.0909...

An example to evaluate normal machine translation outputs

>>> ref1 = str('It is a guide to action that ensures that the military '
...            'will forever heed Party commands').split()
>>> hyp1 = str('It is a guide to action which ensures that the military '
...            'always obeys the commands of the party').split()
>>> hyp2 = str('It is to insure the troops forever hearing the activity '
...            'guidebook that party direct').split()
>>> sentence_gleu([ref1], hyp1) 
0.4393...
>>> sentence_gleu([ref1], hyp2) 
0.1206...

Parameters

• references (list(list(str))) – a list of reference sentences

• hypothesis (list(str)) – a hypothesis sentence

• min_len (int) – The minimum order of n-gram this function should extract.

• max_len (int) – The maximum order of n-gram this function should extract.

Returns

the sentence level GLEU score.

Return type

float