result_string = ""
print("=============================== Filtering data and performing operations ===============================")
# Gets the first letter for the first IPC class and adding it to the ipc_sectons set variable
ipc_sections = set()
for doc in documents:
if len(doc['ipc_classes']) > 0:
ipc_sections.add(doc['ipc_classes'][0][0])
else:
print(doc['filename'])
print(ipc_sections)
ipc_sections = list(ipc_sections)
# Creating a class_map variable, which contains a mapping of the IPC class to a number. The classes are ordered inside
# the classMap method. This mapping is important because of keras library particularities.
class_map = classMap(ipc_sections)
ipc_sections.sort()
embedding_generator = MongoDBMetaEmbeddingGenerator(documents, "section", class_map, len(ipc_sections), serve_forever=True)
print("=============================== Create training classes ===============================")
#Build a factory for a model adapter
model_factory = dl.factory.factory.create('MultilayerKerasRecurrentNN', input_shape=(maxWords, embeddingSize),
numNeurouns=len(ipc_sections), numOutputNeurons=len(ipc_sections), layers=layers)
model = model_factory.create()
model = model.load(model_saved_name)
# Geting the test documents collection
test_documents = mongodb.get_all_meta(testing_documents_collection)
test_embedding_generator = MongoDBMetaEmbeddingGenerator(test_documents, "section", class_map, len(ipc_sections))
result_string = ""
print(
"=============================== Filtering data and performing operations ==============================="
)
# Gets the first letter for the first IPC class and adding it to the ipc_sectons set variable
ipc_sections = set()
for doc in documents:
if len(doc['ipc_classes']) > 0:
ipc_sections.add(doc['ipc_classes'][0][0])
else:
print(doc['filename'])
# Creating a class_map variable, which contains a mapping of the IPC class to a number. The classes are ordered inside
# the classMap method. This mapping is important because of keras library particularities.
class_map = classMap(list(ipc_sections))
# Rebooting mongodb cursor
training_documents = mongodb.get_all_meta(training_documents_collection)
# The Generator for metadata and word embedding, its a python generator that returns "embeding, ipc_class
embedding_generator = MongoDBMetaEmbeddingGenerator(documents,
"section",
class_map,
len(ipc_sections),
serve_forever=True,
to_categorical=False)
print(
"=============================== Create training classes ==============================="
)
#Build a factory for a model adapter
else:
print(doc['filename'])
print(ipc_sections)
print(ipc_classes)
# Creating a class_map variable, which contains a mapping of the IPC class to a number. The classes are ordered inside
# the classMap method. This mapping is important because of keras library particularities.
for key in ipc_classes.keys():
model_saved_name = "../TrainedLSTM/keras_rnn_mongo_float_" + key + ".model"
result_file_name = "../TrainedLSTM/result_rnn_mongo_float_" + key
result_string = ""
classes = ipc_classes[key]
classes.sort()
class_map = classMap(classes)
documents = mongodb.get_meta_by_section(training_documents_collection, key)
# Rebooting mongodb cursor
training_documents = mongodb.get_all_meta(training_documents_collection)
# The Generator for metadata and word embedding, its a python generator that returns "embeding, ipc_class
embedding_generator = MongoDBMetaEmbeddingGenerator(documents,
"class",
class_map,
len(classes),
serve_forever=True)
print("=============================== Create training classes " + key +
" ===============================")
#Build a factory for a model adapter
model_factory = dl.factory.factory.create('MultilayerKerasRecurrentNN',
input_shape=(maxWords,
embeddingSize),