print('INFO: Loading annotated data.')
annot_data = list()
relations = list()
with open(par.a, mode='r', encoding='U8') as f:
for line in f:
lab, parent, child = line.rstrip('\n').split('\t')
relations.append((parent, child))
cust_feat = dict()
if par.r:
cust_feat = dict(custom_features[(parent, child)])
child = child.split('_')
parent = parent.split('_')
lab = True if lab == '+' else False
features = make_vector(parent=parent[0],
parent_pos=parent[1],
child=child[0],
child_pos=child[1],
custom=cust_feat)
annot_data.append({**features, **{'result': lab}})
# split annotated data on train/validation/holdout
divided = split_data(relations,
train=0.65,
validation=0.15,
holdout=0.2,
random_seed=24)
for item in annot_data:
parent = item['parent'] + '_' + item['parentPos']
child = item['child'] + '_' + item['childPos']
item['data'] = divided[(parent, child)]
# load data and assign features
print('INFO: Loading annotated data.')
annot_data = list()
relations = list()
with open(par.a, mode='r', encoding='U8') as f:
for line in f:
lab, parent, child = line.rstrip('\n').split('\t')
relations.append((parent, child))
child = child.split('_')
parent = parent.split('_')
lab = True if lab == '+' else False
features = make_vector(parent=parent[1],
parent_pos=parent[2][0],
child=child[1],
child_pos=child[2][0],
custom={
'Pid': parent[0],
'Cid': child[0]
})
annot_data.append({**features, **{'result': lab}})
# split annotated data on train/validation/holdout
divided = split_data(relations,
train=0.65,
validation=0.15,
holdout=0.2,
random_seed=24)
for item in annot_data:
parent = item['Pid'] + '_' + item['parent'] + '_' + item['parentPos']
child = item['Cid'] + '_' + item['child'] + '_' + item['childPos']
item['data'] = divided[(parent, child)]
'stThPar', 'stFoPar', 'stFiPar', 'enTwPar', 'enThPar',
'enFoPar', 'enFiPar', 'enTwChi', 'enThChi', 'enFoChi',
'enFiChi')
# load manually annotated data
rows_list = list()
with open(par.a, mode='r', encoding='utf-8') as f:
for line in f:
lab, child, parent = line.rstrip('\n').split('\t')
child = child.split('_')
parent = parent.split('_')
lab = 1 if lab == '+' else 0
features = make_vector(parent=parent[0],
parent_pos=parent[1],
child=child[0],
child_pos=child[1])
vector = OrderedDict()
for key in included_features:
vector[key] = features[key]
vector = {**vector, **{'result': lab}}
rows_list.append(vector)
# load data to predict (if given)
if par.p:
with open(par.p, mode='r', encoding='utf-8') as f:
for line in f:
child, parent = line.rstrip('\n').split('\t')
child = child.split('_')