用python代码实现对词频出现次数较高的词汇放入特征词典中，31个类的特征词都放到一个文档里。然后对训练集的31个类的文档用one-hot向量化，即将每个文档中的每个句子表示成与特征词典一样长度的向量，遍历特征词典，只要句子中包含当前的特征词，标记为1，否则标记为0。

可以使用中的`collections.Counter`来实现对词频出现次数较高的词汇的计数，然后将出现次数最高的词放入特征词典中。代码示例如下：

import os
import collections

all_word_counts = collections.Counter()
dict_file = 'feature_dict.txt'

# 统计所有文本的词频
classes = os.listdir('train_data')
for classname in classes:
    class_path = os.path.join('train_data', classname)
    files = os.listdir(class_path)
    for filename in files:
        filepath = os.path.join(class_path, filename)
        with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:
            text = f.read()
            words = text.split()
            all_word_counts.update(words)

# 获取出现次数最高的前N个词
N = 1000
most_common_words = [word for (word, count) in all_word_counts.most_common(N)]

# 将特征词典写入文件
with open(dict_file, 'w', encoding='utf-8') as f:
    f.write('\n'.join(most_common_words))

然后，可以使用以下代码将训练集的文本转换为one-hot向量化的形式：

import numpy as np

# 读取特征词典
with open(dict_file, 'r', encoding='utf-8') as f:
    feature_dict = f.read().split()

# 遍历所有文本，将每个词语转换为one-hot向量
num_classes = 31
class_vectors = []
for classname in range(num_classes):
    class_vectors.append([])
    class_path = os.path.join('train_data', str(classname))
    files = os.listdir(class_path)
    for filename in files:
        filepath = os.path.join(class_path, filename)
        with open(filepath, 'r', encoding='utf-8', errors='ignore') as f:
            text = f.read().lower()
            words = text.split()
            vector = np.zeros(len(feature_dict))
            for word in words:
                if word in feature_dict:
                    index = feature_dict.index(word)
                    vector[index] = 1
            class_vectors[classname].append(vector)

# 将列表转换为numpy数组
class_vectors = [np.array(vec_list) for vec_list in class_vectors]