def create_y(data):
y = []
for doc in data:
for sentence in doc.sentences:
for pair in sentence.pairs:
label = class_index["null"]
if pair.ddi == "true":
label = class_index[pair.type]
y.append(label)
return y
### SPLIT DATASET ###
data = read_dataset()
n_docs = len(data)
np.random.seed(42)
train_amount = 0.7
train_ids = choice(n_docs, int(train_amount * n_docs), replace=False)
test_ids = [i for i in range(n_docs) if i not in train_ids]
training = [data[i] for i in train_ids]
test = [data[i] for i in test_ids]
print("%i training documents" % len(training))
print("%i test documents" % len(test))
### TRAINING ###
#features.append(("drug1_type", drug1_type))
#features.append(("drug2_type", drug2_type))
#drug1_name = pair.e1.text
#drug2_name = pair.e2.text
#features.append(("drug_name", drug1_name))
#features.append(("drug_name", drug2_name))
same_drug = pair.e1.text.lower() == pair.e2.text.lower()
features.append(("same_drug", same_drug))
return features
if __name__ == "__main__":
np.random.seed(42)
data = read_dataset()
n_docs = len(data)
n_folds = 10
folds = k_folds(n_docs, n_folds)
if TEST:
n_folds = 1
classes = ["int", "effect", "none", "mechanism", "advise"]
cv_results = {}
cv_precisions = []
cv_recalls = []
cv_fs = []
cv_2class_precisions = []
cv_2class_recalls = []
cv_2class_fs = []