实战案例：微博情感分析

数据：每个文本文件包含相应类的数据

0：喜悦；1：愤怒；2：厌恶；3：低落

步骤

1. 文本读取
2. 分割训练集、测试集
3. 特征提取
4. 模型训练、预测

代码：

tools.py

# -*- coding: utf-8 -*-

import re
import jieba.posseg as pseg
import pandas as pd
import math
import numpy as np

# 加载常用停用词
stopwords1 = [line.rstrip() for line in open('./中文停用词库.txt', 'r', encoding='utf-8')]
# stopwords2 = [line.rstrip() for line in open('./哈工大停用词表.txt', 'r', encoding='utf-8')]
# stopwords3 = [line.rstrip() for line in open('./四川大学机器智能实验室停用词库.txt', 'r', encoding='utf-8')]
# stopwords = stopwords1 + stopwords2 + stopwords3
stopwords = stopwords1


def proc_text(raw_line):
    """
        处理每行的文本数据
        返回分词结果
    """
    # 1. 使用正则表达式去除非中文字符
    filter_pattern = re.compile('[^\u4E00-\u9FD5]+')
    chinese_only = filter_pattern.sub('', raw_line)

    # 2. 结巴分词+词性标注
    words_lst = pseg.cut(chinese_only)

    # 3. 去除停用词
    meaninful_words = []
    for word, flag in words_lst:
        # if (word not in stopwords) and (flag == 'v'):
            # 也可根据词性去除非动词等
        if word not in stopwords:
            meaninful_words.append(word)

    return ' '.join(meaninful_words)


def split_train_test(text_df, size=0.8):
    """
        分割训练集和测试集
    """
    # 为保证每个类中的数据能在训练集中和测试集中的比例相同，所以需要依次对每个类进行处理
    train_text_df = pd.DataFrame()
    test_text_df = pd.DataFrame()

    labels = [0, 1, 2, 3]
    for label in labels:
        # 找出label的记录
        text_df_w_label = text_df[text_df['label'] == label]
        # 重新设置索引，保证每个类的记录是从0开始索引，方便之后的拆分
        text_df_w_label = text_df_w_label.reset_index()

        # 默认按80%训练集，20%测试集分割
        # 这里为了简化操作，取前80%放到训练集中，后20%放到测试集中
        # 当然也可以随机拆分80%，20%（尝试实现下DataFrame中的随机拆分）

        # 该类数据的行数
        n_lines = text_df_w_label.shape[0]
        split_line_no = math.floor(n_lines * size)
        text_df_w_label_train = text_df_w_label.iloc[:split_line_no, :]
        text_df_w_label_test = text_df_w_label.iloc[split_line_no:, :]

        # 放入整体训练集，测试集中
        train_text_df = train_text_df.append(text_df_w_label_train)
        test_text_df = test_text_df.append(text_df_w_label_test)

    train_text_df = train_text_df.reset_index()
    test_text_df = test_text_df.reset_index()
    return train_text_df, test_text_df


def get_word_list_from_data(text_df):
    """
        将数据集中的单词放入到一个列表中
    """
    word_list = []
    for _, r_data in text_df.iterrows():
        word_list += r_data['text'].split(' ')
    return word_list


def extract_feat_from_data(text_df, text_collection, common_words_freqs):
    """
        特征提取
    """
    # 这里只选择TF-IDF特征作为例子
    # 可考虑使用词频或其他文本特征作为额外的特征

    n_sample = text_df.shape[0]
    n_feat = len(common_words_freqs)
    common_words = [word for word, _ in common_words_freqs]

    # 初始化
    X = np.zeros([n_sample, n_feat])
    y = np.zeros(n_sample)

    print('提取特征...')
    for i, r_data in text_df.iterrows():
        if (i + 1) % 5000 == 0:
            print('已完成{}个样本的特征提取'.format(i + 1))

        text = r_data['text']

        feat_vec = []
        for word in common_words:
            if word in text:
                # 如果在高频词中，计算TF-IDF值
                tf_idf_val = text_collection.tf_idf(word, text)
            else:
                tf_idf_val = 0

            feat_vec.append(tf_idf_val)

        # 赋值
        X[i, :] = np.array(feat_vec)
        y[i] = int(r_data['label'])

    return X, y


def cal_acc(true_labels, pred_labels):
    """
        计算准确率
    """
    n_total = len(true_labels)
    correct_list = [true_labels[i] == pred_labels[i] for i in range(n_total)]

    acc = sum(correct_list) / n_total
    return acc

main.py

# main.py

# -*- coding: utf-8 -*-


import os
import pandas as pd
import nltk
from tools import proc_text, split_train_test, get_word_list_from_data, \
    extract_feat_from_data, cal_acc
from nltk.text import TextCollection
from sklearn.naive_bayes import GaussianNB

dataset_path = './dataset'
text_filenames = ['0_simplifyweibo.txt', '1_simplifyweibo.txt',
                  '2_simplifyweibo.txt', '3_simplifyweibo.txt']

# 原始数据的csv文件
output_text_filename = 'raw_weibo_text.csv'

# 清洗好的文本数据文件
output_cln_text_filename = 'clean_weibo_text.csv'

# 处理和清洗文本数据的时间较长，通过设置is_first_run进行配置
# 如果是第一次运行需要对原始文本数据进行处理和清洗，需要设为True
# 如果之前已经处理了文本数据，并已经保存了清洗好的文本数据，设为False即可
is_first_run = True


def read_and_save_to_csv():
    """
        读取原始文本数据，将标签和文本数据保存成csv
    """

    text_w_label_df_lst = []
    for text_filename in text_filenames:
        text_file = os.path.join(dataset_path, text_filename)

        # 获取标签，即0, 1, 2, 3
        label = int(text_filename[0])

        # 读取文本文件
        with open(text_file, 'r', encoding='utf-8') as f:
            lines = f.read().splitlines()

        labels = [label] * len(lines)

        text_series = pd.Series(lines)
        label_series = pd.Series(labels)

        # 构造dataframe
        text_w_label_df = pd.concat([label_series, text_series], axis=1)
        text_w_label_df_lst.append(text_w_label_df)

    result_df = pd.concat(text_w_label_df_lst, axis=0)

    # 保存成csv文件
    result_df.columns = ['label', 'text']
    result_df.to_csv(os.path.join(dataset_path, output_text_filename),
                     index=None, encoding='utf-8')


def run_main():
    """
        主函数
    """
    # 1. 数据读取，处理，清洗，准备
    if is_first_run:
        print('处理清洗文本数据中...', end=' ')
        # 如果是第一次运行需要对原始文本数据进行处理和清洗

        # 读取原始文本数据，将标签和文本数据保存成csv
        read_and_save_to_csv()

        # 读取处理好的csv文件，构造数据集
        text_df = pd.read_csv(os.path.join(dataset_path, output_text_filename),
                              encoding='utf-8')

        # 处理文本数据
        text_df['text'] = text_df['text'].apply(proc_text)

        # 过滤空字符串
        text_df = text_df[text_df['text'] != '']

        # 保存处理好的文本数据
        text_df.to_csv(os.path.join(dataset_path, output_cln_text_filename),
                       index=None, encoding='utf-8')
        print('完成，并保存结果。')

    # 2. 分割训练集、测试集
    print('加载处理好的文本数据')
    clean_text_df = pd.read_csv(os.path.join(dataset_path, output_cln_text_filename),
                                encoding='utf-8')
    # 分割训练集和测试集
    train_text_df, test_text_df = split_train_test(clean_text_df)
    # 查看训练集测试集基本信息
    print('训练集中各类的数据个数：', train_text_df.groupby('label').size())
    print('测试集中各类的数据个数：', test_text_df.groupby('label').size())

    # 3. 特征提取
    # 计算词频
    n_common_words = 200

    # 将训练集中的单词拿出来统计词频
    print('统计词频...')
    all_words_in_train = get_word_list_from_data(train_text_df)
    fdisk = nltk.FreqDist(all_words_in_train)
    common_words_freqs = fdisk.most_common(n_common_words)
    print('出现最多的{}个词是：'.format(n_common_words))
    for word, count in common_words_freqs:
        print('{}: {}次'.format(word, count))
    print()

    # 在训练集上提取特征
    text_collection = TextCollection(train_text_df['text'].values.tolist())
    print('训练样本提取特征...', end=' ')
    train_X, train_y = extract_feat_from_data(train_text_df, text_collection, common_words_freqs)
    print('完成')
    print()

    print('测试样本提取特征...', end=' ')
    test_X, test_y = extract_feat_from_data(test_text_df, text_collection, common_words_freqs)
    print('完成')

    # 4. 训练模型Naive Bayes
    print('训练模型...', end=' ')
    gnb = GaussianNB()
    gnb.fit(train_X, train_y)
    print('完成')
    print()

    # 5. 预测
    print('测试模型...', end=' ')
    test_pred = gnb.predict(test_X)
    print('完成')

    # 输出准确率
    print('准确率：', cal_acc(test_y, test_pred))

if __name__ == '__main__':
    run_main()