def run_CNN_model(seg_train,
seg_test,
embedding_path,
embed_dim,
cnn_model,
write_predictions=False,
write_No_rel=False,
initial_predictions=None,
final_predictions=None):
"""
Call CNN models
:param seg_test: test segments
:param cnn_model: choose the model
:param seg_train: train segments
:type write_Predictions: write entities and predictions to file
:param initial_predictions: folder to save the initial relation predictions
:param final_predictions: folder to save the final relation predictions
:param write_No_rel: Write the no-relation predictions back to files
:return: None
"""
if cnn_model == 'segment':
model = Model(data_object=seg_train,
data_object_test=seg_test,
segment=True,
test=True,
write_Predictions=write_predictions)
embedding = Embeddings(embedding_path, model, embedding_dim=embed_dim)
seg_cnn = Segment_CNN(model,
embedding,
initial_predictions=initial_predictions,
final_predictions=final_predictions,
write_No_rel=write_No_rel)
elif cnn_model == 'sentence':
model = Model(data_object=seg_train,
data_object_test=seg_test,
test=True,
write_Predictions=write_predictions)
embedding = Embeddings(embedding_path, model, embedding_dim=embed_dim)
single_sent_cnn = Sentence_CNN(model,
embedding,
initial_predictions=initial_predictions,
final_predictions=final_predictions,
write_No_rel=write_No_rel)
else:
# multilabel sentence-CNN does not have the option to write the predictions back
model = Model(data_object=seg_train,
data_object_test=seg_test,
test=True,
multilabel=True)
embedding = Embeddings(embedding_path, model, embedding_dim=embed_dim)
multi_sent_cnn = Sentence_CNN(model, embedding)
preceding_segs='../data/test_segments/preceding_seg',
concept1_segs='../data/test_segments/concept1_seg',
middle_segs='../data/test_segments/middle_seg',
concept2_segs='../data/test_segments/concept2_seg',
succeeding_segs='../data/test_segments/succeeding_seg',
track='../data/test_segments/track').data_object
connection_train = Set_Connection(CSV=True, sentences='../data/segments/sentence_train',
labels='../data/segments/labels_train',
preceding_segs='../data/segments/preceding_seg',
concept1_segs='../data/segments/concept1_seg',
middle_segs='../data/segments/middle_seg',
concept2_segs='../data/segments/concept2_seg',
succeeding_segs='../data/segments/succeeding_seg',
track='../data/segments/track').data_object
model = Model(data_object=connection_train, data_object_test=connection_test, segment=True, test=True, write_Predictions=True)
embedding=Embeddings(embedding_path, model, embedding_dim=int(sys.argv[2]))
if sys.argv[4] is not None and sys.argv[5] is not None:
seg_cnn = Segment_CNN(model, embedding, final_predictions=sys.argv[5], No_Rel=sys.argv[4])
else:
seg_cnn = Segment_CNN(model, embedding)
else:
connection_test = Set_Connection(CSV=True, sentence_only=True, sentences='../data/test_segments/sentence_train',
labels='../data/test_segments/labels_train',
track='../data/test_segments/track' ).data_object
connection_train = Set_Connection(CSV=True, sentence_only=True, sentences='../data/segments/sentence_train',
labels='../data/segments/labels_train',
track='../data/segments/track' ).data_object
import sys
sys.path.append('../')
from RelEx_NN.model import Model
from RelEx_NN.embeddings import Embeddings
from RelEx_NN.cnn import Sentence_CNN
from segment import Set_Connection
# Use pre-trained word embeddings
embedding_path = "../../word_embeddings/glove.6B.200d.txt"
# path to sentence and label CSV files
data = Set_Connection(CSV=True,
sentence_only=True,
sentences='data/sentence_train',
labels='data/labels_train').data_object
model = Model(data, False)
embedding = Embeddings(embedding_path, model)
# train CNN model
sent_cnn = Sentence_CNN(model, embedding, True)
import sys
sys.path.append('../')
from RelEx_NN.model import Model
from RelEx_NN.embeddings import Embeddings
from RelEx_NN.cnn import Sentence_CNN
from segment import Set_Connection
# Use pre-trained word embeddings
embedding_path = "../../word_embeddings/glove.6B.200d.txt"
# path to sentence and label CSV files
data = Set_Connection(CSV=True,
sentence_only=True,
sentences='data/sentence_train',
labels='data/labels_train').data_object
model = Model(data, False, False, True)
embedding = Embeddings(embedding_path, model)
# train CNN model
sent_cnn = Sentence_CNN(model,
embedding,
True,
filters=300,
drop_out=0.1,
filter_conv=3,
optimizer='adam')
import sys
sys.path.append('../')
from RelEx_NN.model import Model
from RelEx_NN.embeddings import Embeddings
from RelEx_NN.cnn import Segment_CNN
from segment import Set_Connection
# Use pre-trained word embeddings
embedding_path = "../../word_embeddings/glove.6B.200d.txt"
# path to sentence and label CSV files
data = Set_Connection(
CSV=True,
sentences='data/sentence_train',
labels='data/labels_train',
preceding_segs='data/segments/preceding_seg',
concept1_segs='data/segments/concept1_seg',
middle_segs='data/segments/middle_seg',
concept2_segs='data/segments/concept2_seg',
succeeding_segs='data/segments/succeeding_seg').data_object
# data = Set_Connection(CSV=True, sentence_only = False, sentences='../data/sentence_train', labels='../data/labels_train',preceding_segs='../data/segments/preceding_seg', concept1_segs='../data/segments/concept1_seg',middle_segs='../data/segments/middle_seg',concept2_segs='../data/segments/concept2_seg', succeeding_segs='../data/segments/succeeding_seg' ).data_object
model = Model(data)
embedding = Embeddings(embedding_path, model)
# train CNN model
seg_cnn = Segment_CNN(model, embedding, True)
embedding_path = "../word_embeddings/glove.6B.300d.txt"
else:
if int(sys.argv[2]) == 200:
embedding_path = "../word_embeddings/patent_w2v.txt"
if sys.argv[3] == 'segment':
data = Set_Connection(CSV=True,
sentences='../data/segments/sentence_train',
labels='../data/segments/labels_train',
preceding_segs='../data/segments/preceding_seg',
concept1_segs='../data/segments/concept1_seg',
middle_segs='../data/segments/middle_seg',
concept2_segs='../data/segments/concept2_seg',
succeeding_segs='../data/segments/succeeding_seg',
track='../data/segments/track').data_object
model = Model(data, segment=True)
# model = Model(data, segment=True, write_Predictions=True)
embedding = Embeddings(embedding_path,
model,
embedding_dim=int(sys.argv[2]))
# if sys.argv[4] and sys.argv[5]:
# seg_cnn = Segment_CNN(model, embedding, True, final_predictions=sys.argv[5], No_Rel=sys.argv[4])
# else:
seg_cnn = Segment_CNN(model, embedding, True)
else:
data = Set_Connection(CSV=True,
sentence_only=True,
sentences='../data/segments/sentence_train',
labels='../data/segments/labels_train',
track='../data/segments/track').data_object