def __init__(self):
Reference.__init__(self, 'German Wiktionary',
'http://de.wiktionary.org/w/api.php', 'wiktionary.pickle')
#TODO(PM) Add a text interface to login
# Open dictionary of noun categories
with open('reference/kategorien.json', 'r') as store:
self.category_dict = json.load(store)
def __init__(self, reference_tokens, negative_value=0.0):
"""
@param reference the reference translation that hypotheses shall be
scored against. Must be an iterable of tokens (any
type).
"""
Reference.__init__(self, reference_tokens)
self.negative_value = negative_value
def __init__(self, reference_tokens, n=4):
"""
@param reference the reference translation that hypotheses shall be
scored against. Must be an iterable of tokens (any
type).
@param n maximum n-gram order to consider.
"""
Reference.__init__(self, reference_tokens)
self.n = n
# preprocess reference
self._reference_length = len(self._reference_tokens)
self._reference_ngrams = self._get_ngrams(self._reference_tokens, self.n)
def __init__(self, reference_tokens, n=4):
"""
@param reference the reference translation that hypotheses shall be
scored against. Must be an iterable of tokens (any
type).
@param n maximum n-gram order to consider.
"""
Reference.__init__(self, reference_tokens)
self.n = n
# preprocess reference
self._reference_length = len(self._reference_tokens)
self._reference_ngrams = self._get_ngrams(self._reference_tokens,
self.n)
def __init__(self, reference_tokens, n=6, beta=1):
"""
@param reference the reference translation that hypotheses shall be
scored against.
@param n maximum character n-gram order to consider.
@param beta algorithm paramater beta (interpolation weight, needs to be > 0).
"""
if beta <= 0:
raise ValueError("Value of beta needs to be larger than zero!")
Reference.__init__(self, reference_tokens)
self.n = n
self.max_order = n
self.beta_squared = beta ** 2
# The paper specifies that whitespace is ignored, but for a training objective,
#it's perhaps better to leave it in. According to the paper, it makes no
#difference in practise for scoring.
self._reference_string = " ".join(reference_tokens).strip()
# Get n-grams from reference:
self._reference_ngrams = self._get_ngrams(self._reference_string, self.n)
def __init__(self, reference_tokens, additional_flags=''):
"""
Computes the TER of a sentence.
:param reference_tokens: the reference translation that hypotheses shall be
scored against. Must be an iterable of tokens (any
/tmp/3420971.ref type).
:param additional_flags: additional TERCOM flags.
"""
self.d = dict(os.environ.copy())
self.d['LANG'] = 'C'
self.extension = str(random.randint(0, 10000000))
self.hyp_filename = "/tmp/" + self.extension + ".hyp"
self.ref_filename = "/tmp/" + self.extension + ".ref"
self.ter_cmd = "bash " + TER_JAR + " -r " + self.ref_filename + " -h " + self.hyp_filename \
+ additional_flags + "| grep TER | awk '{print $3}'"
self.clean_cmd = "rm -f " + self.ref_filename + " " + self.hyp_filename
# Used to guarantee thread safety
self.lock = threading.Lock()
Reference.__init__(self, reference_tokens)
self._gts_ter = ' '.join(reference_tokens) + '\t(sentence%d)\n' % 0
with open(self.ref_filename, 'w') as f:
f.write(self._gts_ter)
def __init__(self, reference_tokens, beer_scorer):
Reference.__init__(self, reference_tokens)
#Construct reference string from tokens
self._reference_string = " ".join(reference_tokens)
self._beer_scorer = beer_scorer
def __init__(self, reference_tokens, meteor_scorer):
Reference.__init__(self, reference_tokens)
#Construct reference string from tokens
self._reference_string = " ".join(reference_tokens)
self._meteor_scorer = meteor_scorer
def __init__(self, outfile, folder=False, opt=False, write_current=False, with_summary=True):
self.with_summary = with_summary
Reference.__init__(self, outfile, folder=folder, opt=opt, write_current=write_current)
