基于rnn神经网络的写唐诗机器人

基于rnn网络的自动写唐诗机器人

项目目录：
my_rnn_model用来保存训练的模型；

clean_txt.py 脚本负责将清洗唐诗集（poetry.txt）

代码如下

poem_dir = "./poetry.txt"
# 读txt文件
with open(poem_dir, "r", encoding="utf-8") as f:
    lines = f.readlines()
    # 定义一个诗集列表用来装诗
    poem_collections = []
    with open("new_poetry.txt", "w", encoding="utf-8") as f1:
        for line in lines:
            # maxsplit=1 是将line分成两份
            if len(line)<10 or len(line)>100:
                continue
            name, content = line.strip().split(":", maxsplit=1)
            poem_collections.append(content)
            for word in name:
            	# 清洗唐诗集中的不规则字符
                if word not in "_()\/:《[] ":
                    f1.write(word)
            f1.write(":")
            for word in content:
                if word not in "_()\/:《[] ":
                    f1.write(word)
            f1.write("\n")

接下来就训练模型的代码 my_poem_train.py:

1.在训练有关文字的模型时，往往需要先构建文字字典，用来包含所有文字和其索引，这样后来就可以构建词向量。代码如下：

# -------------------------------数据预处理---------------------------#
# 诗库目录
poem_dir = "./new_poetry.txt"
# 读txt文件
with open(poem_dir, "r", encoding="utf-8") as f:
    lines =