如何用词向量做文本分类（embedding+cnn）

1、数据简介

本文使用的数据集是著名的”20 Newsgroup dataset”。该数据集共有20种新闻文本数据，我们将实现对该数据集的文本分类任务。数据集的说明和下载请参考（http://www.cs.cmu.edu/afs/cs.cmu.edu/project/theo-20/www/data/news20.html）。本文使用GloVe词向量。GloVe 是 “Global Vectors for Word Representation”的缩写，一种基于共现矩阵分解的词向量。本文所使用的GloVe词向量是在2014年的英文维基百科上训练的，有400k个不同的词，每个词用100维向量表示。链接（http://nlp.stanford.edu/data/glove.6B.zip） (友情提示，词向量文件大小约为822M)

2、数据预处理

我们首先遍历下语料文件下的所有文件夹，获得不同类别的新闻以及对应的类别标签，代码如下所示

texts = []  # list of text sampleslabels_index = {}  # dictionary mapping label name to numeric idlabels = []  # list of label idsimport osTEXT_DATA_DIR = 'e:/textm/20_newsgroup'for name in sorted(os.listdir(TEXT_DATA_DIR)):    path = os.path.join(TEXT_DATA_DIR, name)    if os.path.isdir(path):        label_id = len(labels_index)        labels_index[name] = label_id        if label_id == 2:            break        for fname in sorted(os.listdir(path)):            if fname.isdigit():                fpath = os.path.join(path, fname)                f = open(fpath,'r',encoding='latin-1')                texts.append(f.read().strip())                f.close()                labels.append(label_id)print('Found %s texts.' % len(texts))print(texts[0])print(labels)

之后，我们可以新闻样本转化为神经网络训练所用的张量。所用到的Keras库是keras.preprocessing.text.Tokenizer和keras.preprocessing.sequence.pad_sequences。代码如下所示

######，我们可以新闻样本转化为神经网络训练所用的张量。# 所用到的Keras库是keras.preprocessing.text.Tokenizer和keras.preprocessing.sequence.pad_sequences。代码如下所示from keras.preprocessing.text import Tokenizerfrom keras.preprocessing.sequence import pad_sequencesimport numpy as nptokenizer = Tokenizer()tokenizer.fit_on_texts(texts)sequences = tokenizer.texts_to_sequences(texts)word_index = tokenizer.word_indexprint('Found %s unique tokens.' % len(word_index))data = pad_sequences(sequences)# from keras.utils import np_utils# labels = np_utils.to_categorical(np.asarray(labels))print('Shape of data tensor:', data.shape)# split the data into a training set and a validation setindices = np.arange(data.shape[0])np.random.shuffle(indices)data = data[indices]labels_new = []for i in indices:    labels_new.append(labels[i])nb_validation_samples = int(0.8 * data.shape[0])x_train = data[:-nb_validation_samples]y_train = labels_new[:-nb_validation_samples]x_val = data[-nb_validation_samples:]y_val = labels_new[-nb_validation_samples:]print(x_train[0])

接下来，我们从GloVe文件中解析出每个词和它所对应的词向量，并用字典的方式存储

###############读取词向量embeddings_index = {}f = open(os.path.join('E:\\textm', 'glove.6B.100d.txt'),'r',encoding='utf-8')for line in f.readlines():    values = line.split()    word = values[0]    coefs = np.asarray(values[1:], dtype='float32')    embeddings_index[word] = coefsf.close()print('Found %s word vectors.' % len(embeddings_index))

此时，我们可以根据得到的字典生成上文所定义的词向量矩阵

#############我们可以根据得到的字典生成上文所定义的词向量矩阵embedding_matrix = np.zeros((len(word_index) + 1, 100))for word, i in word_index.items():    embedding_vector = embeddings_index.get(word)    if embedding_vector is not None:        # words not found in embedding index will be all-zeros.        embedding_matrix[i] = embedding_vectorprint(embedding_matrix)#########我们将这个词向量矩阵加载到Embedding层中，注意，我们设置trainable=False使得这个编码层不可再训练。from keras.layers import Embeddingembedding_layer = Embedding(len(word_index) + 1,                            100,                            weights=[embedding_matrix],                            input_length=10036,                            trainable=False)

3、训练模型

用到了三层卷积

from keras.models import *from keras.layers import *from keras.applications import *from keras.preprocessing.image import *sequence_input = Input(shape=(10036,), dtype='int32')embedded_sequences = embedding_layer(sequence_input)x = Conv1D(128, 5, activation='relu')(embedded_sequences)x = MaxPooling1D(5)(x)x = Conv1D(128, 5, activation='relu')(x)x = MaxPooling1D(5)(x)x = Conv1D(128, 5, activation='relu')(x)x = MaxPooling1D(35)(x)  # global max poolingx = Flatten()(x)x = Dense(128, activation='relu')(x)preds = Dense(1, activation='sigmoid')(x)model = Model(sequence_input, preds)model.compile(loss='binary_crossentropy',              optimizer='rmsprop',              metrics=['acc'])# happy learning!model.fit(x_train, y_train, validation_data=(x_val, y_val),          nb_epoch=4, batch_size=128)model.save('e:/mymodel.h5')

全部代码如下

texts = []  # list of text sampleslabels_index = {}  # dictionary mapping label name to numeric idlabels = []  # list of label idsimport osTEXT_DATA_DIR = 'e:/textm/20_newsgroup'for name in sorted(os.listdir(TEXT_DATA_DIR)):    path = os.path.join(TEXT_DATA_DIR, name)    if os.path.isdir(path):        label_id = len(labels_index)        labels_index[name] = label_id        if label_id == 2:            break        for fname in sorted(os.listdir(path)):            if fname.isdigit():                fpath = os.path.join(path, fname)                f = open(fpath,'r',encoding='latin-1')                texts.append(f.read().strip())                f.close()                labels.append(label_id)print('Found %s texts.' % len(texts))print(texts[0])print(labels)######，我们可以新闻样本转化为神经网络训练所用的张量。# 所用到的Keras库是keras.preprocessing.text.Tokenizer和keras.preprocessing.sequence.pad_sequences。代码如下所示from keras.preprocessing.text import Tokenizerfrom keras.preprocessing.sequence import pad_sequencesimport numpy as nptokenizer = Tokenizer()tokenizer.fit_on_texts(texts)sequences = tokenizer.texts_to_sequences(texts)word_index = tokenizer.word_indexprint('Found %s unique tokens.' % len(word_index))data = pad_sequences(sequences)# from keras.utils import np_utils# labels = np_utils.to_categorical(np.asarray(labels))print('Shape of data tensor:', data.shape)# split the data into a training set and a validation setindices = np.arange(data.shape[0])np.random.shuffle(indices)data = data[indices]labels_new = []for i in indices:    labels_new.append(labels[i])nb_validation_samples = int(0.8 * data.shape[0])x_train = data[:-nb_validation_samples]y_train = labels_new[:-nb_validation_samples]x_val = data[-nb_validation_samples:]y_val = labels_new[-nb_validation_samples:]print(x_train[0])###############读取词向量embeddings_index = {}f = open(os.path.join('E:\\textm', 'glove.6B.100d.txt'),'r',encoding='utf-8')for line in f.readlines():    values = line.split()    word = values[0]    coefs = np.asarray(values[1:], dtype='float32')    embeddings_index[word] = coefsf.close()print('Found %s word vectors.' % len(embeddings_index))#############我们可以根据得到的字典生成上文所定义的词向量矩阵embedding_matrix = np.zeros((len(word_index) + 1, 100))for word, i in word_index.items():    embedding_vector = embeddings_index.get(word)    if embedding_vector is not None:        # words not found in embedding index will be all-zeros.        embedding_matrix[i] = embedding_vectorprint(embedding_matrix)#########我们将这个词向量矩阵加载到Embedding层中，注意，我们设置trainable=False使得这个编码层不可再训练。from keras.layers import Embeddingembedding_layer = Embedding(len(word_index) + 1,                            100,                            weights=[embedding_matrix],                            input_length=10036,                            trainable=False)from keras.models import *from keras.layers import *from keras.applications import *from keras.preprocessing.image import *sequence_input = Input(shape=(10036,), dtype='int32')embedded_sequences = embedding_layer(sequence_input)x = Conv1D(128, 5, activation='relu')(embedded_sequences)x = MaxPooling1D(5)(x)x = Conv1D(128, 5, activation='relu')(x)x = MaxPooling1D(5)(x)x = Conv1D(128, 5, activation='relu')(x)x = MaxPooling1D(35)(x)  # global max poolingx = Flatten()(x)x = Dense(128, activation='relu')(x)preds = Dense(1, activation='sigmoid')(x)model = Model(sequence_input, preds)model.compile(loss='binary_crossentropy',              optimizer='rmsprop',              metrics=['acc'])# happy learning!model.fit(x_train, y_train, validation_data=(x_val, y_val),          nb_epoch=4, batch_size=128)model.save('e:/mymodel.h5')

4、参考文献

http://keras-cn.readthedocs.io/en/latest/blog/word_embedding/