Bert深度学习文本分类

哇哦，果不其然，回来之后就开始拖延症了。目前准备两手抓了，一手抓深度学习，一手抓研发那些。再再再次立下flag：一边更新炼丹秘籍，一边更新研发知识。

本次更新有关nlp的知识，其实关于bert的原理的东西，我也不说很熟，不过用来做文本任务，确实很香呀，好用呀！！！

多说无益，可否一学！！！

一、数据EDA

首先来看，本次的动力源-数据！！！

本次需要三个文件，训练文件，测试文件以及标签文件。

train_dataset_path = "train.txt"
test_dataset_path = "test.txt"
label_dataset_path = "class.txt"

train_df = pd.read_csv(train_dataset_path,encoding="utf-8",sep='\t',names=["text","label"])
test_df = pd.read_csv(test_dataset_path,encoding="utf-8",sep = '\t',names=["text","label"])
label_df = pd.read_csv(label_dataset_path,encoding="utf-8",header=None,sep = '\t')

二、数据清洗

数据一般都是不干净的，所以需要对数据清洗一下，让它干干净净的送入炼丹炉中。

首先去除无标签的数据

train_df.dropna(axis=0, how='any', inplace=True)

之后对数据本身进行清洗

import re
def filter(text):
    text = re.sub("[A-Za-z0-9\!\=\？\%\[\]\,\（\）\>\<:&lt;\/#\. -----\_]", "", text)
    text = text.replace('图片', '')
    text = text.replace('\xa0', '') # 删除nbsp
    # new
    r1 =  "\\【.*?】+|\\《.*?》+|\\#.*?#+|[.!/_,$&%^*()<>+""'?@|:~{}#]+|[——！\\\，。=？、：“”‘’￥……（）《》【】]"
    cleanr = re.compile('<.*?>')
    text = re.sub(cleanr, ' ', text)        #去除html标签
    text = re.sub(r1,'',text)
    text = text.strip()
    return text

def clean_text(data):
    data['text'] = data['text'].apply(lambda x: filter(x))
    return data

train = clean_text(train_df)
test = clean_text(test_df)

清洗完数据之后，查看一下数据在各个类别的数量，数据分布很均衡。

三、bert准备工作

要想用bert的话，那么需要安装bert对应的包以及权重文件。本次使用keras来做，当然现在pytorch的transformers也封装了bert，等有机会整理发下。

# 安装keras支持的bert包
!pip install keras_bert

# 下载所需的bert权重参数
!wget -c https://storage.googleapis.com/chineseglue/pretrain_models/roeberta_zh_L-24_H-1024_A-16.zip

四、训练

那么，接下来就将数据喂入模型中，即可开启训练！！！

#! -*- coding:utf-8 -*-
import re, os, json, codecs, gc
import numpy as np
import pandas as pd
from random import choice
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import KFold
from keras_bert import load_trained_model_from_checkpoint, Tokenizer

from keras.layers import *
from keras.callbacks import *
from keras.models import Model
import keras.backend as K
from keras.optimizers import Adam

maxlen = 512
config_path = './bert_config_large.json'
# checkpoint_path = '/export/home/liuyuzhong/kaggle/bert/chinese_L-12_H-768_A-12/bert_model.ckpt'
checkpoint_path = './roberta_zh_large_model.ckpt'
dict_path = './vocab.txt'

token_dict = {}
with codecs.open(dict_path, 'r', 'utf8') as reader:
    for line in reader:
        token = line.strip()
        token_dict[token] = len(token_dict)

class OurTokenizer(Tokenizer):
    def _tokenize(self, text):
        R = []
        for c in text:
            if c in self._token_dict:
                R.append(c)
            elif self._is_space(c):
                R.append('[unused1]') # space类用未经训练的[unused1]表示
            else:
                R.append('[UNK]') # 剩余的字符是[UNK]
        return R

tokenizer = OurTokenizer(token_dict)

def seq_padding(X, padding=0):
    L = [len(x) for x in X]
    ML = max(L)
    return np.array([
        np.concatenate([x, [padding] * (ML - len(x))]) if len(x) < ML else x for x in X
    ])

class data_generator:
    def __init__(self, data, batch_size=8, shuffle=True):
        self.data = data
        self.batch_size = batch_size
        self.shuffle = shuffle
        self.steps = len(self.data) // self.batch_size
        if len(self.data) % self.batch_size != 0:
            self.steps += 1
    def __len__(self):
        return self.steps
    def __iter__(self):
        while True:
            idxs = list(range(len(self.data)))
            
            if self.shuffle:
                np.random.shuffle(idxs)
            
            X1, X2, Y = [], [], []
            for i in idxs:
                d = self.data[i]
                text = d[0][:maxlen]
                x1, x2 = tokenizer.encode(first=text)
                y = d[1]
                X1.append(x1)
                X2.append(x2)
                Y.append([y])
                if len(X1) == self.batch_size or i == idxs[-1]:
                    X1 = seq_padding(X1)
                    X2 = seq_padding(X2)
                    Y = seq_padding(Y)
                    yield [X1, X2], Y[:, 0, :]
                    [X1, X2, Y] = [], [], []

from keras.metrics import top_k_categorical_accuracy
def acc_top5(y_true, y_pred):
    return top_k_categorical_accuracy(y_true, y_pred, k=5)
                    
def build_bert(nclass):
    bert_model = load_trained_model_from_checkpoint(config_path, checkpoint_path, seq_len=None)

    for l in bert_model.layers:
        l.trainable = True

    x1_in = Input(shape=(None,))
    x2_in = Input(shape=(None,))

    x = bert_model([x1_in, x2_in])
    x = Lambda(lambda x: x[:, 0])(x)
    p = Dense(nclass, activation='softmax')(x)

    model = Model([x1_in, x2_in], p)
    model.compile(loss='categorical_crossentropy', 
                  optimizer=Adam(1e-5),
                  metrics=['accuracy', acc_top5])
    print(model.summary())
    return model
    

from keras.utils import to_categorical

DATA_LIST = []
# 改
for data_row in train_df.iloc[:].itertuples():
    DATA_LIST.append((data_row.ocr, to_categorical(data_row.label, 10)))
DATA_LIST = np.array(DATA_LIST)

DATA_LIST_TEST = []
for data_row in test_df.iloc[:].itertuples():
    DATA_LIST_TEST.append((data_row.ocr, to_categorical(0, 10)))
DATA_LIST_TEST = np.array(DATA_LIST_TEST)

def run_cv(nfold, data, data_label, data_test):
    kf = KFold(n_splits=nfold, shuffle=True, random_state=520).split(data)
    # 改
    train_model_pred = np.zeros((len(data), 10))
    test_model_pred = np.zeros((len(data_test), 10))

    for i, (train_fold, test_fold) in enumerate(kf):
        X_train, X_valid, = data[train_fold, :], data[test_fold, :]
        # 改
        model = build_bert(10)
        early_stopping = EarlyStopping(monitor='val_acc', patience=3)
        plateau = ReduceLROnPlateau(monitor="val_acc", verbose=1, mode='max', factor=0.5, patience=2)
        checkpoint = ModelCheckpoint('./bert_dump/' + str(i) + '.hdf5', monitor='val_acc', 
                                         verbose=2, save_best_only=True, mode='max',save_weights_only=True)
        
        train_D = data_generator(X_train, shuffle=True)
        valid_D = data_generator(X_valid, shuffle=True)
        test_D = data_generator(data_test, shuffle=False)
        
        history = model.fit_generator(
            train_D.__iter__(),
            steps_per_epoch=len(train_D),
            epochs=3,
            validation_data=valid_D.__iter__(),
            validation_steps=len(valid_D),
            callbacks=[early_stopping, plateau, checkpoint],
        )
        
        # model.load_weights('./bert_dump/' + str(i) + '.hdf5')
        
        # return model
        train_model_pred[test_fold, :] =  model.predict_generator(valid_D.__iter__(), steps=len(valid_D),verbose=1)
        test_model_pred += model.predict_generator(test_D.__iter__(), steps=len(test_D),verbose=1)
        
        del model; gc.collect()
        K.clear_session()
        
        # break
        
    return train_model_pred, test_model_pred,history

最后的最后，给出本次链接github

那就下篇见！！！