def __init__(self):
self.dep_acc_by_pos = TwoLevelCountDict()
self.head_acc_by_pos = TwoLevelCountDict()
self.long_sent_stats = CountDict()
self.short_sent_stats = CountDict()
self.fields = ["pos_acc", "ul_acc", "l_acc"]
def parse_test(self):
path = '/Users/rgeorgi/Documents/treebanks/LDC95T07/RAW/combined/wsj/00/wsj_0001.mrg'
tc = CountDict()
def count_tokens(tokens):
for token in tokens:
tc.add(token.label)
process_wsj_file(path, count_tokens)
# There should be 31 total tokens in this file.
self.assertEqual(31, tc.total())
self.assertEqual(tc['.'], 2)
class ConllEval(object):
def __init__(self):
self.dep_acc_by_pos = TwoLevelCountDict()
self.head_acc_by_pos = TwoLevelCountDict()
self.long_sent_stats = CountDict()
self.short_sent_stats = CountDict()
self.fields = ["pos_acc", "ul_acc", "l_acc"]
def add(self, k, sent):
self.long_sent_stats.add(k)
if len(sent) < 10:
self.short_sent_stats.add(k)
def pos_stats(self):
for pos in sorted(set(self.dep_acc_by_pos.keys()).union(set(self.head_acc_by_pos.keys()))):
print(
",".join(
[
pos,
str(self.dep_acc_by_pos.sub_distribution(pos).get(True, 0.0)),
str(self.head_acc_by_pos.sub_distribution(pos).get(True, 0.0)),
]
)
)
def acc(self, d, k):
return d[k] / d["words"] * 100
def long_stats(self):
return [self.acc(self.long_sent_stats, k) for k in self.fields]
def short_stats(self):
return [self.acc(self.short_sent_stats, k) for k in self.fields]
def short_ul(self):
return self.acc(self.short_sent_stats, "ul_acc")
def short_ul_count(self):
return self.short_sent_stats.get("ul_acc", 0)
def short_words(self):
return self.short_sent_stats.get("words", 0)
def long_ul(self):
return self.acc(self.long_sent_stats, "ul_acc")
def long_ul_count(self):
return self.long_sent_stats.get("ul_acc", 0)
def long_words(self):
return self.long_sent_stats.get("words", 0)
def __init__(self, gold=None):
CountDict.__init__(self)
self._gold = gold
