def test_hong2014(self):
duc2004 = load_references(_duc2004_file_path)
centroid = load_summaries(_centroid_file_path)
use_porter_stemmer = True
remove_stopwords = False
compute_rouge_l = True
max_words = 100
rouge = Rouge(max_ngram=2,
use_porter_stemmer=use_porter_stemmer,
remove_stopwords=remove_stopwords,
max_words=max_words,
compute_rouge_l=compute_rouge_l)
python_rouge = PythonRouge(use_porter_stemmer=use_porter_stemmer,
remove_stopwords=remove_stopwords,
max_words=max_words,
compute_rouge_l=compute_rouge_l)
expected_metrics, _ = rouge.evaluate(centroid, duc2004)
actual_metrics, _ = python_rouge.evaluate(centroid, duc2004)
assert math.isclose(expected_metrics['rouge-1']['precision'],
actual_metrics['python-rouge-1']['precision'],
abs_tol=1e-2)
assert math.isclose(expected_metrics['rouge-1']['recall'],
actual_metrics['python-rouge-1']['recall'],
abs_tol=2e-2)
assert math.isclose(expected_metrics['rouge-1']['f1'],
actual_metrics['python-rouge-1']['f1'],
abs_tol=2e-2)
assert math.isclose(expected_metrics['rouge-2']['precision'],
actual_metrics['python-rouge-2']['precision'],
abs_tol=1e-2)
assert math.isclose(expected_metrics['rouge-2']['recall'],
actual_metrics['python-rouge-2']['recall'],
abs_tol=1e-2)
assert math.isclose(expected_metrics['rouge-2']['f1'],
actual_metrics['python-rouge-2']['f1'],
abs_tol=1e-2)
# Rouge-L is a little further off, but still reasonably close enough that I'm not worried
assert math.isclose(expected_metrics['rouge-l']['precision'],
actual_metrics['python-rouge-l']['precision'],
abs_tol=1e-1)
assert math.isclose(expected_metrics['rouge-l']['recall'],
actual_metrics['python-rouge-l']['recall'],
abs_tol=1e-1)
assert math.isclose(expected_metrics['rouge-l']['f1'],
actual_metrics['python-rouge-l']['f1'],
abs_tol=1e-1)
def test_python_rouge_multiling(self):
use_porter_stemmer = True
remove_stopwords = False
compute_rouge_l = True
max_words = 100
rouge = Rouge(max_ngram=2,
use_porter_stemmer=use_porter_stemmer,
remove_stopwords=remove_stopwords,
max_words=max_words,
compute_rouge_l=compute_rouge_l)
python_rouge = PythonRouge(use_porter_stemmer=use_porter_stemmer,
remove_stopwords=remove_stopwords,
max_words=max_words,
compute_rouge_l=compute_rouge_l)
expected_metrics, _ = rouge.evaluate(self.summaries,
self.references_list)
actual_metrics, _ = python_rouge.evaluate(self.summaries,
self.references_list)
self.assert_same_as_rouge(actual_metrics, expected_metrics)