Model.py

#coding:utf-8
import numpy as np
import tensorflow as tf

class Model(object):

	def get_config(self):
		return self.config

	def get_positive_instance(self, in_batch = True):
		if in_batch:
			return [self.positive_h, self.positive_t, self.positive_r]
		else:
			return [self.batch_h[0:self.config.batch_size], \
			self.batch_t[0:self.config.batch_size], \
			self.batch_r[0:self.config.batch_size]]

	def get_negative_instance(self, in_batch = True):
		if in_batch:
			return [self.negative_h, self.negative_t, self.negative_r]
		else:
			return [self.batch_h[self.config.batch_size:self.config.batch_seq_size],\
			self.batch_t[self.config.batch_size:self.config.batch_seq_size],\
			self.batch_r[self.config.batch_size:self.config.batch_seq_size]]

	def get_positive_labels(self, in_batch = True):
		if in_batch:
			return self.positive_y
		else:
			return self.batch_y[0:self.config.batch_size]

	def get_negative_labels(self, in_batch = True):
		if in_batch:
			return self.negative_y
		else:
			return self.batch_y[self.config.batch_size:self.config.batch_seq_size]

	def get_all_instance(self, in_batch = False):
		if in_batch:
			return [tf.transpose(tf.reshape(self.batch_h, [1 + self.config.negative_ent + self.config.negative_rel, -1]), [1, 0]),\
			tf.transpose(tf.reshape(self.batch_t, [1 + self.config.negative_ent + self.config.negative_rel, -1]), [1, 0]),\
			tf.transpose(tf.reshape(self.batch_r, [1 + self.config.negative_ent + self.config.negative_rel, -1]), [1, 0])]
		else:
			return [self.batch_h, self.batch_t, self.batch_r]

	def get_all_labels(self, in_batch = False):
		if in_batch:
			return tf.transpose(tf.reshape(self.batch_y, [1 + self.config.negative_ent + self.config.negative_rel, -1]), [1, 0])
		else:
			return self.batch_y

	def get_predict_instance(self):
		return [self.predict_h, self.predict_t, self.predict_r]

	def input_def(self):
		config = self.config
		self.batch_h = tf.placeholder(tf.int64, [config.batch_seq_size])
		self.batch_t = tf.placeholder(tf.int64, [config.batch_seq_size])
		self.batch_r = tf.placeholder(tf.int64, [config.batch_seq_size])
		self.batch_y = tf.placeholder(tf.float32, [config.batch_seq_size])

		self.positive_h = tf.transpose(tf.reshape(self.batch_h[0:config.batch_size], [1, -1]), perm = [1, 0])
		self.positive_t = tf.transpose(tf.reshape(self.batch_t[0:config.batch_size], [1, -1]), perm = [1, 0])
		self.positive_r = tf.transpose(tf.reshape(self.batch_r[0:config.batch_size], [1, -1]), perm = [1, 0])
		self.positive_y = tf.transpose(tf.reshape(self.batch_y[0:config.batch_size], [1, -1]), perm = [1, 0])
		self.negative_h = tf.transpose(tf.reshape(self.batch_h[config.batch_size:config.batch_seq_size], [config.negative_ent + config.negative_rel, -1]), perm = [1, 0])
		self.negative_t = tf.transpose(tf.reshape(self.batch_t[config.batch_size:config.batch_seq_size], [config.negative_ent + config.negative_rel, -1]), perm = [1, 0])
		self.negative_r = tf.transpose(tf.reshape(self.batch_r[config.batch_size:config.batch_seq_size], [config.negative_ent + config.negative_rel, -1]), perm = [1, 0])
		self.negative_y = tf.transpose(tf.reshape(self.batch_y[config.batch_size:config.batch_seq_size], [config.negative_ent + config.negative_rel, -1]), perm = [1, 0])

		self.predict_h = tf.placeholder(tf.int64, [None])
		self.predict_t = tf.placeholder(tf.int64, [None])
		self.predict_r = tf.placeholder(tf.int64, [None])
		self.parameter_lists = []

	def embedding_def(self):
		pass

	def loss_def(self):
		pass

	def predict_def(self):
		pass

	def __init__(self, config, define=False):
		self.config = config

		if define:
			with tf.name_scope("input"):
				self.input_def()

			with tf.name_scope("embedding"):
				self.embedding_def()

			with tf.name_scope("loss"):
				self.loss_def()

			with tf.name_scope("predict"):
				self.predict_def()