def __init__(self, name, data_set=None, answer=None, classifier=None):
if type(name) is not str:
raise TypeError('ClassificationNode name must be str', name)
self.name = name
self.sub_nodes = []
self.classifier = ClassifierBase()
self.parent_node = None
self.data_set = data_set
if answer is not None and type(answer) is not AnswerMessage:
raise ValueError('Answer has to be AnswerMessage or '
'None: [{}] {}'.format(answer, type(answer)))
self.answer = answer
if classifier is not None:
self.set_classifier(classifier)
def _init_classifiers(cls, config, all_features):
'''
Init classifiers by settings.
'''
classifiers = {}
for name, classifier_config in config["classifier_configs"].items():
classifier = ClassifierBase.create_classifier(name, config["feature_extraction_parameters"], classifier_config, all_features)
classifiers[name] = classifier
return classifiers
def _init_classifiers(cls, config, all_features):
'''
Init classifiers by settings.
'''
classifiers = {}
for name, classifier_config in config["classifier_configs"].items():
# create classifiers. TODO how to understand classifier?
# TODO understand feature extraction parameters. Looks like some
# default feature value?
classifier = ClassifierBase.create_classifier(name, config["feature_extraction_parameters"], classifier_config, all_features)
classifiers[name] = classifier
return classifiers
class ClassificationNode(object):
def __init__(self, name, data_set=None, answer=None, classifier=None):
if type(name) is not str:
raise TypeError('ClassificationNode name must be str', name)
self.name = name
self.sub_nodes = []
self.classifier = ClassifierBase()
self.parent_node = None
self.data_set = data_set
if answer is not None and type(answer) is not AnswerMessage:
raise ValueError('Answer has to be AnswerMessage or '
'None: [{}] {}'.format(answer, type(answer)))
self.answer = answer
if classifier is not None:
self.set_classifier(classifier)
def __repr__(self):
res = '<{}> Name: \'{}\' Classifier: \'{}\' Answer: \'{}\''.format(
type(self).__name__, self.name,
type(self.classifier).__name__, self.answer)
# answers = self.classifier.get_answers()
# if answers is not None:
# res += ' Options: \'{}\''.format(answers)
return res
def set_classifier(self, classifier):
if ClassifierBase not in type(classifier).__bases__ and \
type(classifier) is not ClassifierBase:
raise TypeError('Classifier has to be based on ClassifierBase',
classifier)
self.classifier = classifier
def train_classifier(self):
self.classifier.train()
def answer_message(self, message):
if self.answer is None:
raise ValueError('Answer has to be not None in {}'.format(self))
self.answer.options = self.classifier.get_answers()
return self.answer
# if self.answer.function is not None:
# answer = eval(
# 'nodes.external_node_functions.' + self.answer.function
# )(message=message, answer_params=self.answer)
# if type(answer) is AnswerMessage:
# return answer
# return AnswerMessage(text=answer)
# else:
# return self.answer
def parse_message(self, message):
if self.classifier is None:
raise AttributeError(
'Error: classifier is not set.\n{}'.format(self))
node = self.classifier.predict(message=message)
if type(node) is not ClassificationNode:
return None, node
return node, node.answer_message(message)
def add_sub_node(self, sub_node):
if type(sub_node) is not ClassificationNode:
raise TypeError('Sub node must be ClassificationNode')
if self.classifier is None:
raise ValueError('Classifier is None')
self.sub_nodes.append(sub_node)
self.classifier.add_option(sub_node)
sub_node.parent_node = self
def show_children_tree(self,
prefix=' ▶',
first_cycle=True,
full_info=False):
if first_cycle:
print('\n{} children tree:'.format(self.name))
if full_info:
print(prefix, self)
else:
print(prefix, self.name)
for sub in self.sub_nodes:
ClassificationNode.show_children_tree(sub,
prefix='\t' + prefix,
full_info=full_info,
first_cycle=False)
if first_cycle:
print('')
def show_parents_tree(self, full_info=False):
print('\n{} parents tree:'.format(self.name))
if full_info:
sup_tree = [self.__repr__()]
else:
sup_tree = [self.name]
node = self
while node.parent_node is not None:
node = node.parent_node
if full_info:
sup_tree.append(node.__repr__())
else:
sup_tree.append(node.name)
prefix = ' ▶ '
for i in range(len(sup_tree)).__reversed__():
print(prefix + sup_tree[i])
prefix = '\t' + prefix
print('')
def show_full_tree(self, full_info=False):
node = self
while node.parent_node is not None:
node = node.parent_node
node.show_children_tree(full_info=full_info)