fasttext的python实现---

2023-01-31 01:38

短信预约 -IT技能 免费直播动态提醒

转载请注明作者和出处：http://blog.csdn.net/john_bh/

一、简介
二、FastText原理
- 2.1 模型架构
- 2.2 层次SoftMax
- 2.3 N-gram特征
三、基于fastText实现文本分类
- 3.1 fastText有监督学习分类
- 3.2 fastText有监督学习分类
三、总结
- 3.1 fastText和word2vec的区别
- 3.2 小结

fasttext是facebook开源的一个词向量与文本分类工具，在2016年开源，典型应用场景是“带监督的文本分类问题”。提供简单而高效的文本分类和表征学习的方法，性能比肩深度学习而且速度更快。

fastText结合了自然语言处理和机器学习中最成功的理念。这些包括了使用词袋以及n-gram袋表征语句，还有使用子字(subword)信息，并通过隐藏表征在类别间共享信息。我们另外采用了一个softmax层级(利用了类别不均衡分布的优势)来加速运算过程。

这些不同概念被用于两个不同任务：

有效文本分类 ：有监督学习
学习词向量表征：无监督学习

举例来说：fastText能够学会“男孩”、“女孩”、“男人”、“女人”指代的是特定的性别，并且能够将这些数值存在相关文档中。然后，当某个程序在提出一个用户请求（假设是“我女友现在在儿？”），它能够马上在fastText生成的文档中进行查找并且理解用户想要问的是有关女性的问题。

fastText方法包含三部分，模型架构，层次SoftMax和N-gram特征。

2.1 模型架构

fastText的架构和word2vec中的CBOW的架构类似，因为它们的作者都是Facebook的科学家Tomas Mikolov，而且确实fastText也算是words2vec所衍生出来的。

Continuous Bog-Of-Words：

fastText

fastText 模型输入一个词的序列（一段文本或者一句话)，输出这个词序列属于不同类别的概率。
序列中的词和词组组成特征向量，特征向量通过线性变换映射到中间层，中间层再映射到标签。
fastText 在预测标签时使用了非线性激活函数，但在中间层不使用非线性激活函数。

fastText 模型架构和 Word2Vec 中的 CBOW 模型很类似。不同之处在于，fastText 预测标签，而 CBOW 模型预测中间词。

__label__2 , birchas chaim , yeshiva birchas chaim is a orthodox jewish mesivta high school in lakewood township new jersey . it was founded by rabbi shmuel zalmen stein in 2001 after his father rabbi chaim stein asked him to open a branch of telshe yeshiva in lakewood . as of the 2009-10 school year the school had an enrollment of 76 students and 6 . 6 classroom teachers ( on a fte basis ) for a student–teacher ratio of 11 . 5 1 . 
__label__6 , motor torpedo boat pt-41 , motor torpedo boat pt-41 was a pt-20-class motor torpedo boat of the united states navy built by the electric launch company of bayonne new jersey . the boat was laid down as motor boat submarine chaser ptc-21 but was reclassified as pt-41 prior to its launch on 8 july 1941 and was completed on 23 july 1941 . 
__label__11 , passiflora picturata , passiflora picturata is a species of passion flower in the passifloraceae family . 
__label__13 , naya din nai raat , naya din nai raat is a 1974 bollywood drama film directed by a . bhimsingh . the film is famous as sanjeev kumar reprised the nine-role epic performance by sivaji ganesan in navarathri ( 1964 ) which was also previously reprised by akkineni nageswara rao in navarathri ( telugu 1966 ) . this film had enhanced his status and reputation as an actor in hindi cinema .

其中前面的label是前缀，也可以自己定义，label后接的为类别。

具体代码：

# -*- coding:utf-8 -*-
import pandas as pd
import random
import fasttext
import jieba
from sklearn.model_selection import train_test_split

cate_dic = {'technology': 1, 'car': 2, 'entertainment': 3, 'military': 4, 'sports': 5}
"""
函数说明：加载数据
"""
def loadData():


    #利用pandas把数据读进来
    df_technology = pd.read_csv("./data/technology_news.csv",encoding ="utf-8")
    df_technology=df_technology.dropna()    #去空行处理

    df_car = pd.read_csv("./data/car_news.csv",encoding ="utf-8")
    df_car=df_car.dropna()

    df_entertainment = pd.read_csv("./data/entertainment_news.csv",encoding ="utf-8")
    df_entertainment=df_entertainment.dropna()

    df_military = pd.read_csv("./data/military_news.csv",encoding ="utf-8")
    df_military=df_military.dropna()

    df_sports = pd.read_csv("./data/sports_news.csv",encoding ="utf-8")
    df_sports=df_sports.dropna()

    technology=df_technology.content.values.tolist()[1000:21000]
    car=df_car.content.values.tolist()[1000:21000]
    entertainment=df_entertainment.content.values.tolist()[:20000]
    military=df_military.content.values.tolist()[:20000]
    sports=df_sports.content.values.tolist()[:20000]

    return technology,car,entertainment,military,sports

"""
函数说明：停用词
参数说明：
    datapath：停用词路径
返回值：
    stopwords:停用词
"""
def getStopWords(datapath):
    stopwords=pd.read_csv(datapath,index_col=False,quoting=3,sep="\t",names=['stopword'], encoding='utf-8')
    stopwords=stopwords["stopword"].values
    return stopwords

"""
函数说明：去停用词
参数：
    content_line：文本数据
    sentences：存储的数据
    category：文本类别
"""
def preprocess_text(content_line,sentences,category,stopwords):
    for line in content_line:
        try:
            segs=jieba.lcut(line)    #利用结巴分词进行中文分词
            segs=filter(lambda x:len(x)>1,segs)    #去掉长度小于1的词
            segs=filter(lambda x:x not in stopwords,segs)    #去掉停用词
            sentences.append("__lable__"+str(category)+" , "+" ".join(segs))    #把当前的文本和对应的类别拼接起来，组合成fasttext的文本格式
        except Exception as e:
            print (line)
            continue

"""
函数说明：把处理好的写入到文件中，备用
参数说明：

"""
def writeData(sentences,fileName):
    print("writing data to fasttext format...")
    out=open(fileName,'w')
    for sentence in sentences:
        out.write(sentence.encode('utf8')+"\n")
    print("done!")

"""
函数说明：数据处理
"""
def preprocessData(stopwords,saveDataFile):
    technology,car,entertainment,military,sports=loadData()    

    #去停用词，生成数据集
    sentences=[]
    preprocess_text(technology,sentences,cate_dic["technology"],stopwords)
    preprocess_text(car,sentences,cate_dic["car"],stopwords)
    preprocess_text(entertainment,sentences,cate_dic["entertainment"],stopwords)
    preprocess_text(military,sentences,cate_dic["military"],stopwords)
    preprocess_text(sports,sentences,cate_dic["sports"],stopwords)

    random.shuffle(sentences)    #做乱序处理，使得同类别的样本不至于扎堆

    writeData(sentences,saveDataFile)

if __name__=="__main__":
    stopwordsFile=r"./data/stopwords.txt"
    stopwords=getStopWords(stopwordsFile)
    saveDataFile=r'train_data.txt'
    preprocessData(stopwords,saveDataFile)
    #fasttext.supervised():有监督的学习
    classifier=fasttext.supervised(saveDataFile,'classifier.model',lable_prefix='__lable__')
    result = classifier.test(saveDataFile)
    print("P@1:",result.precision)    #准确率
    print("R@2:",result.recall)    #召回率
    print("Number of examples:",result.nexamples)    #预测错的例子

    #实际预测
    lable_to_cate={1:'technology'.1:'car',3:'entertainment',4:'military',5:'sports'}

    texts=['中新网 日电 2018 预赛 亚洲区 强赛 中国队 韩国队 较量 比赛 上半场 分钟 主场 作战 中国队 率先 打破 场上 僵局 利用 角球 机会 大宝 前点 攻门 得手 中国队 领先']
    lables=classifier.predict(texts)
    print(lables)
    print(lable_to_cate[int(lables[0][0])])

    #还可以得到类别+概率
    lables=classifier.predict_proba(texts)
    print(lables)

    #还可以得到前k个类别
    lables=classifier.predict(texts，k=3)
    print(lables)

    #还可以得到前k个类别+概率
    lables=classifier.predict_proba(texts，k=3)
    print(lables)

3.2 fastText有监督学习分类

# -*- coding:utf-8 -*-
import pandas as pd
import random
import fasttext
import jieba
from sklearn.model_selection import train_test_split

cate_dic = {'technology': 1, 'car': 2, 'entertainment': 3, 'military': 4, 'sports': 5}
"""
函数说明：加载数据
"""
def loadData():


    #利用pandas把数据读进来
    df_technology = pd.read_csv("./data/technology_news.csv",encoding ="utf-8")
    df_technology=df_technology.dropna()    #去空行处理

    df_car = pd.read_csv("./data/car_news.csv",encoding ="utf-8")
    df_car=df_car.dropna()

    df_entertainment = pd.read_csv("./data/entertainment_news.csv",encoding ="utf-8")
    df_entertainment=df_entertainment.dropna()

    df_military = pd.read_csv("./data/military_news.csv",encoding ="utf-8")
    df_military=df_military.dropna()

    df_sports = pd.read_csv("./data/sports_news.csv",encoding ="utf-8")
    df_sports=df_sports.dropna()

    technology=df_technology.content.values.tolist()[1000:21000]
    car=df_car.content.values.tolist()[1000:21000]
    entertainment=df_entertainment.content.values.tolist()[:20000]
    military=df_military.content.values.tolist()[:20000]
    sports=df_sports.content.values.tolist()[:20000]

    return technology,car,entertainment,military,sports

"""
函数说明：停用词
参数说明：
    datapath：停用词路径
返回值：
    stopwords:停用词
"""
def getStopWords(datapath):
    stopwords=pd.read_csv(datapath,index_col=False,quoting=3,sep="\t",names=['stopword'], encoding='utf-8')
    stopwords=stopwords["stopword"].values
    return stopwords

"""
函数说明：去停用词
参数：
    content_line：文本数据
    sentences：存储的数据
    category：文本类别
"""
def preprocess_text(content_line,sentences,stopwords):
    for line in content_line:
        try:
            segs=jieba.lcut(line)    #利用结巴分词进行中文分词
            segs=filter(lambda x:len(x)>1,segs)    #去掉长度小于1的词
            segs=filter(lambda x:x not in stopwords,segs)    #去掉停用词
            sentences.append(" ".join(segs))
        except Exception as e:
            print (line)
            continue

"""
函数说明：把处理好的写入到文件中，备用
参数说明：

"""
def writeData(sentences,fileName):
    print("writing data to fasttext format...")
    out=open(fileName,'w')
    for sentence in sentences:
        out.write(sentence.encode('utf8')+"\n")
    print("done!")

"""
函数说明：数据处理
"""
def preprocessData(stopwords,saveDataFile):
    technology,car,entertainment,military,sports=loadData()    

    #去停用词，生成数据集
    sentences=[]
    preprocess_text(technology,sentences,stopwords)
    preprocess_text(car,sentences,stopwords)
    preprocess_text(entertainment,sentences,stopwords)
    preprocess_text(military,sentences,stopwords)
    preprocess_text(sports,sentences,stopwords)

    random.shuffle(sentences)    #做乱序处理，使得同类别的样本不至于扎堆

    writeData(sentences,saveDataFile)


if __name__=="__main__":
    stopwordsFile=r"./data/stopwords.txt"
    stopwords=getStopWords(stopwordsFile)
    saveDataFile=r'unsupervised_train_data.txt'
    preprocessData(stopwords,saveDataFile)

    #fasttext.load_model:不管是有监督还是无监督的，都是载入一个模型
    #fasttext.skipgram(),fasttext.cbow()都是无监督的，用来训练词向量的

    model=fasttext.skipgram('unsupervised_train_data.txt','model')
    print(model.words)    #打印词向量

    #cbow model
    model=fasttext.cbow('unsupervised_train_data.txt','model')
    print(model.words)    #打印词向量