def get_gold_standard_categorization(self):
gold_standard_categorization = {}
categories = self.__get_all_category_directory_names()
for category in categories:
category_path = self.__get_category_path(category)
all_documents_in_category = dataset_handler.get_names_of_files_in_directory(category_path)
category_index_term = n_gram_handler.string_to_index_term(category)
gold_standard_categorization[category_index_term] = all_documents_in_category
return gold_standard_categorization
def get_all_test_documents(self):
all_test_documents = {}
sub_directories = dataset_handler.get_all_subdirectory_names(self.data_path)
for sub_directory in sub_directories:
sub_directory_path = os.path.join(self.data_path, sub_directory)
files_indices = dataset_handler.get_names_of_files_in_directory(sub_directory_path)
for file_id in files_indices:
file_path = os.path.join(sub_directory_path, file_id)
all_test_documents[file_id] = dataset_handler.get_document_as_string(file_path,self.encoding)
return all_test_documents
