Update disrpt.py
Browse files
disrpt.py
CHANGED
@@ -71,23 +71,39 @@ files = [
|
|
71 |
"zho.rst.sctb",
|
72 |
]
|
73 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
|
75 |
def parse_conll_stream(file_stream):
|
76 |
names = ['id', 'form', 'lemma', 'upos', 'xpos', 'feats', 'head', 'deprel', 'deps', 'misc']
|
77 |
sentence = {name: [] for name in names}
|
78 |
-
|
79 |
for line in file_stream:
|
80 |
line = line.strip()
|
81 |
if line.startswith("#"):
|
82 |
continue
|
83 |
if not line:
|
84 |
if sentence['id']:
|
85 |
-
yield sentence
|
86 |
sentence = {name: [] for name in names}
|
87 |
continue
|
88 |
token_data = line.split('\t')
|
89 |
for name, value in zip(names, token_data):
|
90 |
-
|
|
|
|
|
|
|
91 |
|
92 |
def get_kwarg_names(func):
|
93 |
return [k for k, v in inspect.signature(func).parameters.items() if v.default != v.empty]
|
@@ -96,7 +112,7 @@ _URLs = {f'{task}-{split}.{type}':f"https://raw.githubusercontent.com/disrpt/sha
|
|
96 |
for task in files for split in 'train dev test'.split() for type in ['rels','conllu']}
|
97 |
#_URLs = {k:v for k,v in _URLs.items() if requests.get(v).status_code!=404}
|
98 |
|
99 |
-
conllu_features = ['id', 'form', 'lemma', 'upos', 'xpos', 'feats', 'head', 'deprel', 'deps', 'misc', 'seg']
|
100 |
feature_type = {"seg":datasets.features.Sequence(
|
101 |
datasets.features.ClassLabel(names=["O","B-Segment"])),
|
102 |
'id':datasets.Value("string")}
|
|
|
71 |
"zho.rst.sctb",
|
72 |
]
|
73 |
|
74 |
+
def fix_mwe(sentence):
|
75 |
+
mwe={}
|
76 |
+
for i, x in enumerate(sentence['id']):
|
77 |
+
if '-' in x:
|
78 |
+
for a in x.split('-'):
|
79 |
+
mwe[a]=sentence['form'][i]
|
80 |
+
sentence['parent_mwe']+=[mwe.get(x,'')]
|
81 |
+
|
82 |
+
for i, x in enumerate(sentence['id']):
|
83 |
+
if "-" in x:
|
84 |
+
for k,v in sentence.items():
|
85 |
+
del v[i]
|
86 |
+
return sentence
|
87 |
|
88 |
def parse_conll_stream(file_stream):
|
89 |
names = ['id', 'form', 'lemma', 'upos', 'xpos', 'feats', 'head', 'deprel', 'deps', 'misc']
|
90 |
sentence = {name: [] for name in names}
|
91 |
+
mwe_id=[]
|
92 |
for line in file_stream:
|
93 |
line = line.strip()
|
94 |
if line.startswith("#"):
|
95 |
continue
|
96 |
if not line:
|
97 |
if sentence['id']:
|
98 |
+
yield fix_mwe(sentence)
|
99 |
sentence = {name: [] for name in names}
|
100 |
continue
|
101 |
token_data = line.split('\t')
|
102 |
for name, value in zip(names, token_data):
|
103 |
+
if name=='id' and '-' in value:
|
104 |
+
mwe_id=value.split('-')
|
105 |
+
else:
|
106 |
+
sentence[name].append(value)
|
107 |
|
108 |
def get_kwarg_names(func):
|
109 |
return [k for k, v in inspect.signature(func).parameters.items() if v.default != v.empty]
|
|
|
112 |
for task in files for split in 'train dev test'.split() for type in ['rels','conllu']}
|
113 |
#_URLs = {k:v for k,v in _URLs.items() if requests.get(v).status_code!=404}
|
114 |
|
115 |
+
conllu_features = ['id', 'form', 'lemma', 'upos', 'xpos', 'feats', 'head', 'deprel', 'deps', 'misc', 'seg','parent_mwe']
|
116 |
feature_type = {"seg":datasets.features.Sequence(
|
117 |
datasets.features.ClassLabel(names=["O","B-Segment"])),
|
118 |
'id':datasets.Value("string")}
|