def get_evalb_summary(text):
long = {}
short = {}
d = None
for line in text.splitlines():
line = line.strip()
# these lines tell us which dictionary to fill
if line == '-- All --':
d = long
elif line.startswith('-- len'): # e.g. "-- len<40 --"
d = short
# no dictionary yet? no problem!
if d is None:
continue
# we have a dictionary and a line, so try to parse it
try:
k, v = line.split('=')
k = k.strip()
v = try_parse_float(v)
d[k] = v
except:
pass # TODO: warn?
return long, short
def per_sentence_scores(stringiter):
for line in stringiter:
# attempt to clean up length "unmatch"es
match = length_unmatch_re.search(line)
if match:
line = length_unmatch_re.sub('', line).rstrip() + ' '
line += stringiter.next().lstrip()
match = sentence_stats_re.match(line)
if match:
d = {}
for k, v in match.groupdict().items():
d[k] = try_parse_float(v)
d['precision'], d['recall'] = \
calc_precision_recall(d['matchedbracket'],
d['goldbracket'],
d['testbracket'])
d['precision'] *= 100
d['recall'] *= 100
d['fscore'] = calc_fscore(d['precision'], d['recall'])
yield d