使用字符串训练Keras / Tensorflow模型

时间:2020-04-30 21:50:57

标签: python python-3.x string tensorflow deep-learning

我正在尝试找出如何以某种方式转换字符串,以便它们可以
可用于训练聊天机器人的tensorflow / keras模型。

现在我正在使用此方法:

import tensorflow as tf
import numpy as np

messages = open("messages.txt","r").readlines()
responses = open("responses.txt","r").readlines()

train_messages = []
for line in messages:
  tmp1 = []
  for word in line.split(' '):
    tmp2 = []
    for char in word:
      tmp2.append(ord(char))
    tmp1.append(tmp2)
  train_messages.append(tmp1)
train_messages = np.array(train_messages, dtype=int)

train_responses = []
for line in responses:
  tmp1 = []
  for word in line.split(' '):
    tmp2 = []
    for char in word:
      tmp2.append(ord(char))
    tmp1.append(tmp2)
  train_responses.append(tmp1)
train_messages = np.array(train_messages, dtype=int)

# rest of code...

我需要某种方式将整个单词转换为int,我意识到我拥有的代码可能会导致ai拼错许多单词。

我正在Ubuntu Linux上使用官方的Python 3.5发行版。

0 个答案:

没有答案