def testRougeLNonConsecutive(self):
scorer = rouge_scorer.RougeScorer(["rougeL"])
result = scorer.score("testing one two", "testing two")
self.assertAlmostEqual(1, result["rougeL"].precision)
self.assertAlmostEqual(2 / 3, result["rougeL"].recall)
self.assertAlmostEqual(4 / 5, result["rougeL"].fmeasure)
def testInvalidRougeTypes(self, rouge_type):
with self.assertRaises(ValueError):
scorer = rouge_scorer.RougeScorer([rouge_type])
scorer.score("testing one two", "testing")
def testRougeEmpty(self, rouge_type):
scorer = rouge_scorer.RougeScorer([rouge_type])
result = scorer.score("testing one two", "")
self.assertAlmostEqual(0, result[rouge_type].precision)
self.assertAlmostEqual(0, result[rouge_type].recall)
self.assertAlmostEqual(0, result[rouge_type].fmeasure)
def testRouge2(self):
scorer = rouge_scorer.RougeScorer(["rouge2"])
result = scorer.score("testing one two", "testing one")
self.assertAlmostEqual(1, result["rouge2"].precision)
self.assertAlmostEqual(1 / 2, result["rouge2"].recall)
self.assertAlmostEqual(2 / 3, result["rouge2"].fmeasure)
def testValidRougeTypes(self, rouge_type):
scorer = rouge_scorer.RougeScorer([rouge_type])
result = scorer.score("testing one two", "testing")
self.assertSameElements(list(result.keys()), [rouge_type])
from rouge import rouge_scorer
from rouge import scoring
from summae import p2s_eval
from summae import util
FLAGS = flags.FLAGS
flags.DEFINE_string('data_dir', '.', 'Data directory.')
flags.DEFINE_string('eval_subset', 'test',
'which subset (valid/test) to eval/decode.')
flags.DEFINE_string('output_dir', '/tmp/12342',
'local directory to save extractive oracle')
flags.DEFINE_string('vocab_file', '',
'Subword vocab file.') # for detok first sentence
my_rouge_scorer = rouge_scorer.RougeScorer(['rouge1', 'rouge2', 'rougeL'],
use_stemmer=True)
def get_extracts(s):
# get 5 sentences as the extractive baselines
sents = s.feature_lists.feature_list['untokenized_sentences'].feature
assert len(sents) == 5
return tuple([sents[i].bytes_list.value[0] for i in range(5)])
def human_ave(summ_list):
"""Average pairwise rouge between two human summaries."""
agg = scoring.BootstrapAggregator()
for s1_id, s1 in enumerate(summ_list):
for s2_id, s2 in enumerate(summ_list):
if s1_id >= s2_id: # only compute for s1_id < s2_id
def testAssertsOnInvalidInputFiles(self):
scorer = rouge_scorer.RougeScorer(["rouge1"], False)
with self.assertRaises(ValueError):
io.compute_scores_and_write_to_csv("invalid*", "invalid*",
"invalid", scorer,
scoring.BootstrapAggregator())
def testRouge1Multi(self):
scorer = rouge_scorer.RougeScorer(["rouge1"])
result = scorer.score_multi(["testing one two"], "testing")
self.assertAlmostEqual(1, result["rouge1"].precision)
self.assertAlmostEqual(1 / 3, result["rouge1"].recall)
self.assertAlmostEqual(1 / 2, result["rouge1"].fmeasure)
def __init__(self):
self.metrics = ['rouge1', 'rouge2', 'rougeL']
self.main_metric = 'rougeL'
self.scorer = rouge_scorer.RougeScorer(self.metrics, use_stemmer=True)
