计算Slef-BLEU分数python代码

转载

Automated metrics for evaluating the quality of text generation

代码

import numpy as np
import copy
from nltk.translate.bleu_score import sentence_bleu

def get_bleu_score(sentence, remaining_sentences):
    lst = []
    for i in remaining_sentences:
        bleu = sentence_bleu(sentence, i)
        lst.append(bleu)
    return lst


def calculate_selfBleu(sentences):
    '''
    sentences - list of sentences generated by NLG system
    '''
    bleu_scores = []
	
    for i in sentences:
        sentences_copy = copy.deepcopy(sentences)
        remaining_sentences = sentences_copy.remove(i)
        print(sentences_copy)
        bleu = get_bleu_score(i,sentences_copy)
        bleu_scores.append(bleu)

    return np.mean(bleu_scores)

calculate_selfBleu(sentences)