NER实体识别
2019-05-16 本文已影响73人
潇萧之炎
main
#-*-encoding=utf8-*-
from flask import jsonify # 轻量级flask部署用的
from flask import Flask
from flask import request
import json
import platform
import codecs
import logging
import itertools
from collections import OrderedDict # 字数、词频的统计
import os # 目录文件增删改查
import sys # 打断点，获取路径
from gevent import monkey # 优化flask，部署线程优化
monkey.patch_all()
from gevent import wsgi
import tensorflow as tf
import numpy as np
from model import Model
from loader import load_sentences, update_tag_scheme
from loader import char_mapping, tag_mapping #char_mapping, tag_mapping字符映射，tag映射，都是预处理
from loader import augment_with_pretrained, prepare_dataset #预训练，做准备
from utils import get_logger,load_config,create_model #通过create_model加载 Model里面的模型
from utils import make_path
from data_utils import load_word2vec, create_input, input_from_line, BatchManager
currentPath=os.getcwd() # 当前目录
sys.path.append(currentPath)
root_path=os.getcwd()#获取的是根目录
#判断py的版本，获取不同的包
global pyversion
if sys.version>'3':
pyversion='three'
else:
pyversion='two'
if pyversion=='three':
import pickle
else :
import cPickle,pickle
root_path=os.getcwd()+os.sep #os.getcwd()获取当前文件夹的根路径，os.sep是分隔符，转义
flags = tf.app.flags # 用tf.app.flags来定义参数，可以在flags里保存参数
# flags后面之间填写参数或者是文件、文件夹名称
flags.DEFINE_boolean("clean", True, "clean train folder")#清理之前的训练结果
flags.DEFINE_boolean("train", True, "Whether train the model") #是否训练模型
# configurations for the model
flags.DEFINE_integer("seg_dim", 20, "Embedding size for segmentation, 0 if not used")# embeding的增维
# 因为Y是双标签，所以x也要用双标签来标注。BIOS是标注y的，不是x
# 文字有两重信息:1.文字本身的100字向量 2.位置信息：20维
# ，急性呼吸道感染
# 0 1 2 2 2 2 2 3 逗号是0，开头是1，结尾是3，中间全是2
# 比如x急是0100四维，全连接20维，再加上原来的100维，100+20=120维。20就是做位置词的Embedding，用120维来代替一个x的输入
flags.DEFINE_integer("char_dim", 100, "Embedding size for characters")#字的维度
flags.DEFINE_integer("lstm_dim", 100, "Num of hidden units in LSTM, or num of filters in IDCNN")#隐层
flags.DEFINE_string("tag_schema", "iobes", "tagging schema iobes or iob") #y有实体信息和位置信息，这里是标签的位置类型
# configurations for training
flags.DEFINE_float("clip", 5, "Gradient clip")#梯度截断值
flags.DEFINE_float("dropout", 0.5, "Dropout rate")
flags.DEFINE_float("batch_size", 20, "batch size")
flags.DEFINE_float("lr", 0.001, "Initial learning rate")
flags.DEFINE_string("optimizer", "adam", "Optimizer for training")#优化器，tf有9类优化器
flags.DEFINE_boolean("pre_emb", True, "Wither use pre-trained embedding")#数据预处理embeding,char_dim是100，这里就是true
flags.DEFINE_boolean("zeros", True, "Wither replace digits with zero")#碰到生僻字用0取代,预测值为0？
flags.DEFINE_boolean("lower", False, "Wither lower case") # 是否需要将字母小写，这个案例中，字符串不需要小写
flags.DEFINE_integer("max_epoch", 100, "maximum training epochs")# 最大epoch,建议5000-10000
flags.DEFINE_integer("steps_check", 100, "steps per checkpoint")# 每100个batch输出损失
flags.DEFINE_string("ckpt_path", "ckpt", "Path to save model") #保存模型的路径
flags.DEFINE_string("summary_path", "summary", "Path to store summaries")# 保存可视化摘要，保存流程图
flags.DEFINE_string("log_file", "train.log", "File for log") #maps.pkl一般用来保存模型，这里保存字典
flags.DEFINE_string("map_file", "maps.pkl", "file for maps")#保存字典的向量，训练集的正反向字典，将训练集隐射成word2vec的字典
flags.DEFINE_string("vocab_file", "vocab.json", "File for vocab")#原始ccorpus
flags.DEFINE_string("config_file", "config_file", "File for config")#配置文件
flags.DEFINE_string("script", "conlleval", "evaluation script")
flags.DEFINE_string("result_path", "result", "Path for results")
flags.DEFINE_string("emb_file", os.path.join(root_path+"data", "vec.txt"), "Path for pre_trained embedding")
flags.DEFINE_string("train_file", os.path.join(root_path+"data", "example.train"), "Path for train data")#训练集
flags.DEFINE_string("dev_file", os.path.join(root_path+"data", "example.dev"), "Path for dev data")# 开发集或验证集，验证当前模型的损失是否在减小
flags.DEFINE_string("test_file", os.path.join(root_path+"data", "example.test"), "Path for test data")#测试集
#深度学习中，样本一般分为3份，一边训练，一边用验证集中的数据，验证当前模型的损失是否在减小，精确度是否很高，因为量非常大
#该项目中放在D:\Python\NERuselocal\NERuselocal\data文件夹下
flags.DEFINE_string("model_type", "idcnn", "Model type, can be idcnn or bilstm")
#flags.DEFINE_string("model_type", "bilstm", "Model type, can be idcnn or bilstm")
FLAGS = tf.app.flags.FLAGS #上面的参数保存在这里
# 断言，相当于if else的判断 return
assert FLAGS.clip < 5.1, "gradient clip should't be too much"
assert 0 <= FLAGS.dropout < 1, "dropout rate between 0 and 1"
assert FLAGS.lr > 0, "learning rate must larger than zero"
assert FLAGS.optimizer in ["adam", "sgd", "adagrad"]
# config for the model 无调用
def config_model(char_to_id, tag_to_id):
config = OrderedDict()
config["model_type"] = FLAGS.model_type
config["num_chars"] = len(char_to_id)
config["char_dim"] = FLAGS.char_dim
config["num_tags"] = len(tag_to_id)
config["seg_dim"] = FLAGS.seg_dim
config["lstm_dim"] = FLAGS.lstm_dim
config["batch_size"] = FLAGS.batch_size
config["emb_file"] = FLAGS.emb_file
config["clip"] = FLAGS.clip
config["dropout_keep"] = 1.0 - FLAGS.dropout
config["optimizer"] = FLAGS.optimizer
config["lr"] = FLAGS.lr
config["tag_schema"] = FLAGS.tag_schema
config["pre_emb"] = FLAGS.pre_emb
config["zeros"] = FLAGS.zeros
config["lower"] = FLAGS.lower
return config
# 打断点的时候，上面的def config_model（）不执行
# 把x和y的正反向字典读进来了
with open(FLAGS.map_file, "rb") as f:
if pyversion=='three':
#词到ID,标记到ID。pickle用来打开D:\Python\NERuselocal\NERuselocal\maps.pkl文件
# 这四个值就是正反向字典，长度分别是
# 2678 2678 51 51
char_to_id, id_to_char, tag_to_id, id_to_tag = pickle.load(f)
else:
char_to_id, id_to_char, tag_to_id, id_to_tag = pickle.load(f,protocol=2)
# make path for store log and model if not exist
make_path(FLAGS)
if os.path.isfile(FLAGS.config_file):
# 如果D:\Python\NERuselocal\NERuselocal\config_file是个文件，就加载进来
config = load_config(FLAGS.config_file)
else:
config = config_model(char_to_id, tag_to_id)
save_config(config, FLAGS.config_file)
# 全部加载到flag中去
make_path(FLAGS)
app = Flask(__name__)
log_path = os.path.join("log", FLAGS.log_file)
logger = get_logger(log_path)
tf_config = tf.ConfigProto()
sess=tf.Session(config=tf_config)
#sess.run(tf.global_variables_initializer())
# 会话创建好之后加载模型
# id_to_char就是反向字典
model = create_model(sess, Model, FLAGS.ckpt_path, load_word2vec, config, id_to_char, logger)
'''
def get_text_input():
#http://127.0.0.1:5002/?inputStr="最开心"
text=request.args.get('inputStr')
if len(text.strip())>0:
aa=model.evaluate_line(sess, input_from_line(text, char_to_id), id_to_tag)
return jsonify(aa)
'''
@app.route('/', methods=['POST','GET'])
def get_text_input():
#http://127.0.0.1:5002/?inputStr="神经病"
#如果遇到显示问题：下载QQ浏览器，将编码设置为utf-8
text=request.args.get('inputStr')
#if len(text.strip())>0:
if text:
aa=model.evaluate_line(sess, input_from_line(text, char_to_id), id_to_tag)
print(aa)
return jsonify(aa)
@app.route('/text',methods=['POST','GET'])
def text():
#http://101.224.189.118:5002/text?inputStr="神经病"
# 101.224.189.118
text=request.args.get('inputStr')
#if len(text.strip())>0:
#Input_from_line get_seg_features self.trans(crf_log_likelihood) run_step self.loss
#project_layer_idcnn
if text:
aa=model.evaluate_line(sess, input_from_line(text, char_to_id), id_to_tag)
print(aa)
return jsonify(aa)
if __name__ == "__main__":
app.config['JSON_AS_ASCII'] = False
app.run(host='127.0.0.1',port=5002)
# id_to_tag.txt和id_to_tag.txt是正反向字典
# vec.txt是词向量，每个都是100维的，即时是一个逗号。
# 与聊天机器人不同的地方，之前每个字是转化为数值，这里是转化为向量
# main里面放的是测试的结果，main2是训练的过程
# 只要不是函数的，就会从上往下去运行，运行到define就会跳过
mian2
# encoding=utf8
import codecs
import pickle
import itertools
from collections import OrderedDict
import os
import tensorflow as tf
import numpy as np
from model import Model
from loader import load_sentences, update_tag_scheme
from loader import char_mapping, tag_mapping
from loader import augment_with_pretrained, prepare_dataset
from utils import get_logger, make_path, clean, create_model, save_model
from utils import print_config, save_config, load_config, test_ner
from data_utils import load_word2vec, create_input, input_from_line, BatchManager
root_path=os.getcwd()+os.sep
flags = tf.app.flags
flags.DEFINE_boolean("clean", True, "clean train folder") #清理之前的训练结果
flags.DEFINE_boolean("train", False, "Whether train the model") #是否训练模型
# configurations for the model
flags.DEFINE_integer("seg_dim", 20, "Embedding size for segmentation, 0 if not used") # embeding的增维
flags.DEFINE_integer("char_dim", 100, "Embedding size for characters")#字的维度
flags.DEFINE_integer("lstm_dim", 100, "Num of hidden units in LSTM, or num of filters in IDCNN")
flags.DEFINE_string("tag_schema", "iobes", "tagging schema iobes or iob") #标签的位置类型
# configurations for training
flags.DEFINE_float("clip", 5, "Gradient clip") #梯度截断值
flags.DEFINE_float("dropout", 0.5, "Dropout rate")
flags.DEFINE_float("batch_size", 20, "batch size")
flags.DEFINE_float("lr", 0.001, "Initial learning rate")
flags.DEFINE_string("optimizer", "adam", "Optimizer for training")
flags.DEFINE_boolean("pre_emb", True, "Wither use pre-trained embedding")
flags.DEFINE_boolean("zeros", True, "Wither replace digits with zero")
flags.DEFINE_boolean("lower", False, "Wither lower case")
flags.DEFINE_integer("max_epoch", 100, "maximum training epochs") # 建议5000-10000
flags.DEFINE_integer("steps_check", 100, "steps per checkpoint")
flags.DEFINE_string("ckpt_path", "ckpt", "Path to save model") #保存模型的路径
flags.DEFINE_string("summary_path", "summary", "Path to store summaries") # 保存可视化摘要
flags.DEFINE_string("log_file", "train.log", "File for log") #日志
flags.DEFINE_string("map_file", "maps.pkl", "file for maps") #保存字典的向量
flags.DEFINE_string("vocab_file", "vocab.json", "File for vocab")
flags.DEFINE_string("config_file", "config_file", "File for config")
flags.DEFINE_string("script", "conlleval", "evaluation script")
flags.DEFINE_string("result_path", "result", "Path for results")
flags.DEFINE_string("emb_file", os.path.join(root_path+"data", "vec.txt"), "Path for pre_trained embedding")
flags.DEFINE_string("train_file", os.path.join(root_path+"data", "example.train"), "Path for train data")
flags.DEFINE_string("dev_file", os.path.join(root_path+"data", "example.dev"), "Path for dev data") # 验证集，验证是否损失在下降
flags.DEFINE_string("test_file", os.path.join(root_path+"data", "example.test"), "Path for test data")
flags.DEFINE_string("model_type", "idcnn", "Model type, can be idcnn or bilstm")
#flags.DEFINE_string("model_type", "bilstm", "Model type, can be idcnn or bilstm")
FLAGS = tf.app.flags.FLAGS
assert FLAGS.clip < 5.1, "gradient clip should't be too much"
assert 0 <= FLAGS.dropout < 1, "dropout rate between 0 and 1"
assert FLAGS.lr > 0, "learning rate must larger than zero"
assert FLAGS.optimizer in ["adam", "sgd", "adagrad"]
# config for the model
def config_model(char_to_id, tag_to_id):
config = OrderedDict()
config["model_type"] = FLAGS.model_type
config["num_chars"] = len(char_to_id)
config["char_dim"] = FLAGS.char_dim
config["num_tags"] = len(tag_to_id)
config["seg_dim"] = FLAGS.seg_dim
config["lstm_dim"] = FLAGS.lstm_dim
config["batch_size"] = FLAGS.batch_size
config["emb_file"] = FLAGS.emb_file
config["clip"] = FLAGS.clip
config["dropout_keep"] = 1.0 - FLAGS.dropout
config["optimizer"] = FLAGS.optimizer
config["lr"] = FLAGS.lr
config["tag_schema"] = FLAGS.tag_schema
config["pre_emb"] = FLAGS.pre_emb
config["zeros"] = FLAGS.zeros
config["lower"] = FLAGS.lower
return config
def evaluate(sess, model, name, data, id_to_tag, logger):
logger.info("evaluate:{}".format(name))
ner_results = model.evaluate(sess, data, id_to_tag)
eval_lines = test_ner(ner_results, FLAGS.result_path)
for line in eval_lines:
logger.info(line)
f1 = float(eval_lines[1].strip().split()[-1])
if name == "dev":
best_test_f1 = model.best_dev_f1.eval()
if f1 > best_test_f1:
tf.assign(model.best_dev_f1, f1).eval()
logger.info("new best dev f1 score:{:>.3f}".format(f1))
return f1 > best_test_f1
elif name == "test":
best_test_f1 = model.best_test_f1.eval()
if f1 > best_test_f1:
tf.assign(model.best_test_f1, f1).eval()
logger.info("new best test f1 score:{:>.3f}".format(f1))
return f1 > best_test_f1
def train():
# load data sets [class_list['[', 'O'], ['双', 'O'], ['击', 'O']]…] 详细讲 data/example.*
train_sentences = load_sentences(FLAGS.train_file, FLAGS.lower, FLAGS.zeros)
dev_sentences = load_sentences(FLAGS.dev_file, FLAGS.lower, FLAGS.zeros)
test_sentences = load_sentences(FLAGS.test_file, FLAGS.lower, FLAGS.zeros)
# Use selected tagging scheme (IOB / IOBES) 详细讲 IOB：实体开头为I,其他为I,非实体O
update_tag_scheme(train_sentences, FLAGS.tag_schema)
update_tag_scheme(test_sentences, FLAGS.tag_schema)
update_tag_scheme(dev_sentences, FLAGS.tag_schema)
# create maps if not exist
if not os.path.isfile(FLAGS.map_file):
# create dictionary for word
if FLAGS.pre_emb:
dico_chars_train = char_mapping(train_sentences, FLAGS.lower)[0]
dico_chars, char_to_id, id_to_char = augment_with_pretrained(
dico_chars_train.copy(),
FLAGS.emb_file,
list(itertools.chain.from_iterable(
[[w[0] for w in s] for s in test_sentences])
)
)
else:
_c, char_to_id, id_to_char = char_mapping(train_sentences, FLAGS.lower)
# Create a dictionary and a mapping for tags
_t, tag_to_id, id_to_tag = tag_mapping(train_sentences)
#with open('maps.txt','w',encoding='utf8') as f1:
#f1.writelines(str(char_to_id)+" "+id_to_char+" "+str(tag_to_id)+" "+id_to_tag+'\n')
with open(FLAGS.map_file, "wb") as f:
pickle.dump([char_to_id, id_to_char, tag_to_id, id_to_tag], f)
else:
with open(FLAGS.map_file, "rb") as f:
char_to_id, id_to_char, tag_to_id, id_to_tag = pickle.load(f)
# prepare data, get a collection of list containing index
train_data = prepare_dataset(
train_sentences, char_to_id, tag_to_id, FLAGS.lower
)
dev_data = prepare_dataset(
dev_sentences, char_to_id, tag_to_id, FLAGS.lower
)
test_data = prepare_dataset(
test_sentences, char_to_id, tag_to_id, FLAGS.lower
)
print("%i / %i / %i sentences in train / dev / test." % (
len(train_data), 0, len(test_data)))
train_manager = BatchManager(train_data, FLAGS.batch_size)
dev_manager = BatchManager(dev_data, 100)
test_manager = BatchManager(test_data, 100)
# make path for store log and model if not exist
make_path(FLAGS)
if os.path.isfile(FLAGS.config_file):
config = load_config(FLAGS.config_file)
else:
config = config_model(char_to_id, tag_to_id)
save_config(config, FLAGS.config_file)
make_path(FLAGS)
log_path = os.path.join("log", FLAGS.log_file)
logger = get_logger(log_path)
print_config(config, logger)
# limit GPU memory
tf_config = tf.ConfigProto()
tf_config.gpu_options.allow_growth = True
steps_per_epoch = train_manager.len_data
with tf.Session(config=tf_config) as sess:
model = create_model(sess, Model, FLAGS.ckpt_path, load_word2vec, config, id_to_char, logger)
logger.info("start training")
loss = []
with tf.device("/gpu:0"):
for i in range(100):
for batch in train_manager.iter_batch(shuffle=True):
step, batch_loss = model.run_step(sess, True, batch)
loss.append(batch_loss)
if step % FLAGS.steps_check == 0:
iteration = step // steps_per_epoch + 1
logger.info("iteration:{} step:{}/{}, "
"NER loss:{:>9.6f}".format(
iteration, step%steps_per_epoch, steps_per_epoch, np.mean(loss)))
loss = []
# best = evaluate(sess, model, "dev", dev_manager, id_to_tag, logger)
if i%7==0:
save_model(sess, model, FLAGS.ckpt_path, logger)
#evaluate(sess, model, "test", test_manager, id_to_tag, logger)
def evaluate_line():
config = load_config(FLAGS.config_file)
logger = get_logger(FLAGS.log_file)
# limit GPU memory
tf_config = tf.ConfigProto()
tf_config.gpu_options.allow_growth = True
with open(FLAGS.map_file, "rb") as f:
char_to_id, id_to_char, tag_to_id, id_to_tag = pickle.load(f)
with tf.Session(config=tf_config) as sess:
model = create_model(sess, Model, FLAGS.ckpt_path, load_word2vec, config, id_to_char, logger)
while True:
# try:
# line = input("请输入测试句子:")
# result = model.evaluate_line(sess, input_from_line(line, char_to_id), id_to_tag)
# print(result)
# except Exception as e:
# logger.info(e)
line = input("请输入测试句子:")
result = model.evaluate_line(sess, input_from_line(line, char_to_id), id_to_tag)
print(result)
def main(_):
if 1:
if FLAGS.clean:
clean(FLAGS)
train()
else:
evaluate_line()
if __name__ == "__main__":
tf.app.run(main)
model
# encoding = utf8
import numpy as np
import tensorflow as tf
from tensorflow.contrib.crf import crf_log_likelihood #条件随机场
from tensorflow.contrib.crf import viterbi_decode # 隐马尔科夫链
from tensorflow.contrib.layers.python.layers import initializers #初始化神经网络
from utils import result_to_json #保存参数
from data_utils import create_input, iobes_iob,iob_iobes
# 双向lstm或IdCNN模型，找到x,y. y是双标签，x是文字word2vec映射成的词向量。
# 如何拟合x.y:拟合之前第一步提取x的特征，用BiLstm或idCNN对x做特征提取，+分类器（crf条件随机场）
# BiLstm or idCNN + crf
# idCNN与cnn的区别是，idCNN的卷积核是扁的：找一句话之间的关系可以用扁的，
# 好处：可以有效地抗噪音：完形填空时，扁的卷积核它只会扫当前这句话，不会把上下文卷进来，抗的是上下文的躁
# CNN和RNN本质上没有太大差别，都是把局部的相关性体现出来，CNN体现在空间上，RNN体现在时间时序上
# crf：条件随机场。跟rnn很类似，提供了一个分类结果，当然它也可以做特征提取。它的分类需要算一个联合概率
# 第一步，找到x,y
# 第二步，对x做特征提取、特征工程（之前所有的resnet等都是为特征工程服务的），对y做one_hot向量（或二分类）
# 第三步，去拟合，分类
# crf_log_likelihood(#likelihood似然，一般加似然的就是损失函数
class Model(object):
def __init__(self, config):
#__init__方法下面的参数都会被执行，相当于构造方法
self.config = config
self.lr = config["lr"]
self.char_dim = config["char_dim"] # embeding_size 100
self.lstm_dim = config["lstm_dim"] # lstm隐层神经元个数
self.seg_dim = config["seg_dim"] #增加的维度
self.num_tags = config["num_tags"] #tag的标签个数
self.num_chars = config["num_chars"] # 字典维度
self.num_segs = 4 #0，1，2，3，0是不需要的字，1是第一个，2是中间的，3是最后一个
self.global_step = tf.Variable(0, trainable=False)
self.best_dev_f1 = tf.Variable(0.0, trainable=False)
self.best_test_f1 = tf.Variable(0.0, trainable=False)
#xavier_initializer迭代器，效率高，和global_initializer类似
self.initializer = initializers.xavier_initializer()
# add placeholders for the model
# batch_size是20
self.char_inputs = tf.placeholder(dtype=tf.int32, # 这个是20*100
shape=[None, None],
name="ChatInputs")
self.seg_inputs = tf.placeholder(dtype=tf.int32, # 这个是20*20，0-3映射成20
shape=[None, None], # 后面加起来120*20
name="SegInputs")
self.targets = tf.placeholder(dtype=tf.int32, # 这个是20*1,y值
shape=[None, None],
name="Targets")
# dropout keep prob
self.dropout = tf.placeholder(dtype=tf.float32,
name="Dropout")
used = tf.sign(tf.abs(self.char_inputs))
length = tf.reduce_sum(used, reduction_indices=1)
#二维的东西，降掉一维，算整个长度是多少
self.lengths = tf.cast(length, tf.int32)# 120
self.batch_size = tf.shape(self.char_inputs)[0] # 20*120，第0个就是20
self.num_steps = tf.shape(self.char_inputs)[-1] # 120，最后一个就是120
#Add model type by crownpku bilstm or idcnn
self.model_type = config['model_type']#idcnn
#parameters for idcnn
# idcnn后面连的是膨胀卷积，好处：有些图像比较小的时候，不希望挤到一起。防止欠拟合
# 一种方法是，把图像做膨胀，另一种方法是将卷积核做膨胀。一般是feature_map做膨胀，卷积核不膨胀
# 由3*3变成5*5，中间补0
self.layers = [
{
'dilation': 1 #膨胀卷积 膨胀卷积核尺寸 = 膨胀系数*（原始卷积核尺寸-1）+1
},
{
'dilation': 1
},
{
'dilation': 2
},
]
self.filter_width = 3 #卷积核宽3,卷积核的高没有写，所以高是1，1*3，卷积核是扁的
self.num_filter = self.lstm_dim #卷积核个数即为lstm连接隐层的个数，就是卷积的通道数输出的
#字向量的维度+词长度特征维度
self.embedding_dim = self.char_dim + self.seg_dim # embedding_size 120=100+20
self.repeat_times = 4 #重复的次数是4，4层卷积网络 深度3*4=12层，重复的是self.layers
self.cnn_output_width = 0 #输出的宽度实际上是2000多维，这里初始化为0
# embeddings for chinese character and segmentation representation
embedding = self.embedding_layer(self.char_inputs, self.seg_inputs, config)
if self.model_type == 'bilstm':
# apply dropout before feed to lstm layer
model_inputs = tf.nn.dropout(embedding, self.dropout)
# bi-directional lstm layer
model_outputs = self.biLSTM_layer(model_inputs, self.lstm_dim, self.lengths)
# logits for tags
self.logits = self.project_layer_bilstm(model_outputs)
elif self.model_type == 'idcnn':
# apply dropout before feed to idcnn layer
# 120个里面随机删掉一部分，内存不删，删里面的值
# dropout在输入层、输出层、隐层都可以做
model_inputs = tf.nn.dropout(embedding, self.dropout) #输入120个
# ldcnn layer
# ldcnn layer 特征提取 膨胀卷积
# model_inputs是120维的，如果做了dropout，就剩60维了
model_outputs = self.IDCNN_layer(model_inputs) #输出200个
# 输入（100+20）个——卷积--》（100个通道---》3次膨胀（100）以上循环4次）
# logits for tags
# logits for tags 模型文件的输出
self.logits = self.project_layer_idcnn(model_outputs)
else:
raise KeyError
# loss of the model
self.loss = self.loss_layer(self.logits, self.lengths)
with tf.variable_scope("optimizer"):
optimizer = self.config["optimizer"]
if optimizer == "sgd":
self.opt = tf.train.GradientDescentOptimizer(self.lr)
elif optimizer == "adam":
self.opt = tf.train.AdamOptimizer(self.lr)
elif optimizer == "adgrad":
self.opt = tf.train.AdagradOptimizer(self.lr)
else:
raise KeyError
# apply grad clip to avoid gradient explosion
grads_vars = self.opt.compute_gradients(self.loss)
capped_grads_vars = [[tf.clip_by_value(g, -self.config["clip"], self.config["clip"]), v]
for g, v in grads_vars]
self.train_op = self.opt.apply_gradients(capped_grads_vars, self.global_step)
# saver of the model
self.saver = tf.train.Saver(tf.global_variables(), max_to_keep=5)
# 用来做embedding。seg_inputs是20维的。
def embedding_layer(self, char_inputs, seg_inputs, config, name=None):
"""
:param char_inputs: one-hot encoding of sentence
:param seg_inputs: segmentation feature
:param config: wither use segmentation feature
:return: [1, num_steps, embedding size],
"""
#高：3 血：22 糖：23 和：24 高：3 血：22 压：25 char_inputs=[3,22,23,24,3,22,25]
#高血糖 和 高血压 seg_inputs 高血糖=[1,2,3] 和=[0] 高血压=[1,2,3] seg_inputs=[1,2,3,0,1,2,3]
embedding = []
with tf.variable_scope("char_embedding" if not name else name), tf.device('/cpu:0'):
# self.num_chars=3538， self.char_dim=100维 ，char_lookup字符查找
self.char_lookup = tf.get_variable(
name="char_embedding",
shape=[self.num_chars, self.char_dim],
initializer=self.initializer)
#输入char_inputs='常'对应的字典的索引/编号/value为：8
#self.char_lookup=[2677*100]的向量，char_inputs字对应在字典的索引/编号/key=[1]
# char_lookup：被查的字典。char_inputs：每个字的索引
# 查的过程是向量相乘，看平板截图
embedding.append(tf.nn.embedding_lookup(self.char_lookup, char_inputs))#把input映射成embedding
#上一步完成后，变成7*100
if config["seg_dim"]:#上面是创建100维的，这里再20维的
#self.num_segs=4, self.seg_dim=20 ，4*20的
with tf.variable_scope("seg_embedding"), tf.device('/cpu:0'):
self.seg_lookup = tf.get_variable(
name="seg_embedding",
shape=[self.num_segs, self.seg_dim],
initializer=self.initializer)
embedding.append(tf.nn.embedding_lookup(self.seg_lookup, seg_inputs))#分割部位的embedding,生成20维
# seg_input只有四个值，0、1、2、3
embed = tf.concat(embedding, axis=-1)#组成120维向量
return embed
def biLSTM_layer(self, model_inputs, lstm_dim, lengths, name=None):
"""
:param lstm_inputs: [batch_size, num_steps, emb_size]
:return: [batch_size, num_steps, 2*lstm_dim]
"""
with tf.variable_scope("char_BiLSTM" if not name else name):
lstm_cell = {}
for direction in ["forward", "backward"]:
with tf.variable_scope(direction):
lstm_cell[direction] = rnn.CoupledInputForgetGateLSTMCell(
lstm_dim,
use_peepholes=True,
initializer=self.initializer,
state_is_tuple=True)
outputs, final_states = tf.nn.bidirectional_dynamic_rnn(
lstm_cell["forward"],
lstm_cell["backward"],
model_inputs,
dtype=tf.float32,
sequence_length=lengths)
return tf.concat(outputs, axis=2)
#IDCNN layer
def IDCNN_layer(self, model_inputs,
name=None):
"""
:param idcnn_inputs: [batch_size, num_steps, emb_size]
:return: [batch_size, num_steps, cnn_output_width]
"""
#ft.expand_dims会向tensor中插入一个维度，插入位置就是参数代表的位置(维度从0开始)
model_inputs = tf.expand_dims(model_inputs, 1) #增加了一个维度
# shape由[?,?,120]变成[?,1,?,120]，最后一维是embedding
reuse = False
if self.dropout == 1.0:
reuse = True
with tf.variable_scope("idcnn" if not name else name):
shape=[1, self.filter_width, self.embedding_dim,
self.num_filter]
print(shape)
filter_weights = tf.get_variable(
"idcnn_filter",
shape=[1, self.filter_width, self.embedding_dim,
self.num_filter],
initializer=self.initializer)
"""
shape of input = [batch, in_height, in_width, in_channels]
shape of filter = [filter_height, filter_width, in_channels, out_channels]
height是默认1，width是句子长度，通道是120维
shape of input = [batch, in_height, in_width, in_channels]
shape of filter = [filter_height, filter_width, in_channels, out_channels]
"""
layerInput = tf.nn.conv2d(model_inputs,
filter_weights,
strides=[1, 1, 1, 1],
padding="SAME",
name="init_layer",use_cudnn_on_gpu=True)
finalOutFromLayers = []
totalWidthForLastDim = 0
#多次卷积，就会将膨胀的时候单次没有卷到的数据在下次卷到
for j in range(self.repeat_times):
for i in range(len(self.layers)):
dilation = self.layers[i]['dilation']
isLast = True if i == (len(self.layers) - 1) else False
with tf.variable_scope("atrous-conv-layer-%d" % i,
reuse=True
if (reuse or j > 0) else False):
w = tf.get_variable(
"filterW",
shape=[1, self.filter_width, self.num_filter,
self.num_filter],
initializer=tf.contrib.layers.xavier_initializer())
b = tf.get_variable("filterB", shape=[self.num_filter])
#膨胀卷积：插入rate-1个0 这里三层{1,1,2}相当于前两个没有膨胀
conv = tf.nn.atrous_conv2d(layerInput,
w,
rate=dilation,
padding="SAME")
conv = tf.nn.bias_add(conv, b)
conv = tf.nn.relu(conv)
if isLast:
finalOutFromLayers.append(conv)
totalWidthForLastDim += self.num_filter
layerInput = conv
finalOut = tf.concat(axis=3, values=finalOutFromLayers)
keepProb = 1.0 if reuse else 0.5
finalOut = tf.nn.dropout(finalOut, keepProb)
#踢掉指定的维度，值不变
finalOut = tf.squeeze(finalOut, [1])
finalOut = tf.reshape(finalOut, [-1, totalWidthForLastDim])
self.cnn_output_width = totalWidthForLastDim
return finalOut
def project_layer_bilstm(self, lstm_outputs, name=None):
"""
hidden layer between lstm layer and logits
:param lstm_outputs: [batch_size, num_steps, emb_size]
:return: [batch_size, num_steps, num_tags]
"""
with tf.variable_scope("project" if not name else name):
with tf.variable_scope("hidden"):
W = tf.get_variable("W", shape=[self.lstm_dim*2, self.lstm_dim],
dtype=tf.float32, initializer=self.initializer)
b = tf.get_variable("b", shape=[self.lstm_dim], dtype=tf.float32,
initializer=tf.zeros_initializer())
output = tf.reshape(lstm_outputs, shape=[-1, self.lstm_dim*2])
hidden = tf.tanh(tf.nn.xw_plus_b(output, W, b))
# project to score of tags
with tf.variable_scope("logits"):
W = tf.get_variable("W", shape=[self.lstm_dim, self.num_tags],
dtype=tf.float32, initializer=self.initializer)
b = tf.get_variable("b", shape=[self.num_tags], dtype=tf.float32,
initializer=tf.zeros_initializer())
pred = tf.nn.xw_plus_b(hidden, W, b)
return tf.reshape(pred, [-1, self.num_steps, self.num_tags])
#Project layer for idcnn by crownpku
#Delete the hidden layer, and change bias initializer
def project_layer_idcnn(self, idcnn_outputs, name=None):
"""
:param lstm_outputs: [batch_size, num_steps, emb_size]
:return: [batch_size, num_steps, num_tags]
"""
with tf.variable_scope("project" if not name else name):
# project to score of tags
with tf.variable_scope("logits"):
W = tf.get_variable("W", shape=[self.cnn_output_width, self.num_tags],
dtype=tf.float32, initializer=self.initializer)
b = tf.get_variable("b", initializer=tf.constant(0.001, shape=[self.num_tags]))
pred = tf.nn.xw_plus_b(idcnn_outputs, W, b)
return tf.reshape(pred, [-1, self.num_steps, self.num_tags])
def loss_layer(self, project_logits, lengths, name=None):
"""
calculate crf loss
:param project_logits: [1, num_steps, num_tags]
:return: scalar loss
"""
#num_steps是句子长度；project_logits是特征提取并全连接后的输出
with tf.variable_scope("crf_loss" if not name else name):
small = -1000.0
# pad logits for crf loss #start_logits=[batch_size,1,num_tags+1]
start_logits = tf.concat(
[small * tf.ones(shape=[self.batch_size, 1, self.num_tags]), tf.zeros(shape=[self.batch_size, 1, 1])], axis=-1)
#pad_logits=[batch_size,num_steps,1]
pad_logits = tf.cast(small * tf.ones([self.batch_size, self.num_steps, 1]), tf.float32)
#logits=[batch_size,num_steps,num_tags+1]
logits = tf.concat([project_logits, pad_logits], axis=-1)
#logits=[batch_size,num_steps+1,num_tags+1]
logits = tf.concat([start_logits, logits], axis=1)
targets = tf.concat(
[tf.cast(self.num_tags*tf.ones([self.batch_size, 1]), tf.int32), self.targets], axis=-1)
#targets=[batch_size,1+实际标签数]
self.trans = tf.get_variable(
"transitions",
shape=[self.num_tags + 1, self.num_tags + 1],
initializer=self.initializer)
#logits是模型的特征输出；targets是label;trans是条件随机场的输出
#crf_log_likelihood在一个条件随机场里计算标签序列的log-likelihood
#inputs:一个形状为[batch_size,max_seq_len,num_tags]的tensor
#一般使用BILSTM处理之后输出转换为他要求的形状作为CRF层的输入
#tag_indices:一个形状为[batch_size]的向量，表示每个序列的长度
#sequence_lengths:一个形状为[batch_size]的向量，表示每个序列的长度
#transition_params:形状为[num_tags,num_tags]的转移矩阵
#log_likelihood:标量，log-likelihood
#注意：由于条件随机场有标记，故真实维度+1
#inputs=[char_inputs,seg_inputs]
#高：3 血：22 糖：23 和：24 高：3 血：22 压：25 char_inputs=[3,22,23,24,3,22,25]
#高血糖 和 高血压 seg_inputs 高血糖=[1,2,3] 和=[0] 高血压=[1,2,3] seg_inputs=[1,2,3,0,1,2,3]
log_likelihood, self.trans = crf_log_likelihood(#likelihood似然，一般加似然的就是损失函数
inputs=logits,
tag_indices=targets,
transition_params=self.trans,
sequence_lengths=lengths+1)
return tf.reduce_mean(-log_likelihood)
def create_feed_dict(self, is_train, batch):
"""
:param is_train: Flag, True for train batch
:param batch: list train/evaluate data
:return: structured data to feed
"""
_, chars, segs, tags = batch
feed_dict = {
self.char_inputs: np.asarray(chars),
self.seg_inputs: np.asarray(segs),
self.dropout: 1.0,
}
if is_train:
feed_dict[self.targets] = np.asarray(tags)
feed_dict[self.dropout] = self.config["dropout_keep"]
return feed_dict
def run_step(self, sess, is_train, batch):
"""
:param sess: session to run the batch
:param is_train: a flag indicate if it is a train batch
:param batch: a dict containing batch data
:return: batch result, loss of the batch or logits
"""
feed_dict = self.create_feed_dict(is_train, batch)
if is_train:
global_step, loss, _ = sess.run(
[self.global_step, self.loss, self.train_op],
feed_dict)
return global_step, loss
else:
#lengths是字的个数，logits是模型特征
lengths, logits = sess.run([self.lengths, self.logits], feed_dict)
return lengths, logits
def decode(self, logits, lengths, matrix):
"""
:param logits: [batch_size, num_steps, num_tags]float32, logits
:param lengths: [batch_size]int32, real length of each sequence
:param matrix: transaction matrix for inference
:return:
"""
# inference final labels usa viterbi Algorithm
paths = []
small = -1000.0
start = np.asarray([[small]*self.num_tags +[0]])
for score, length in zip(logits, lengths):
score = score[:length]
pad = small * np.ones([length, 1])
logits = np.concatenate([score, pad], axis=1)
logits = np.concatenate([start, logits], axis=0)
#由显式序列logits和状态转移阵matrix，求隐藏序列的最大概率路径，也即最短路径
path, _ = viterbi_decode(logits, matrix)
paths.append(path[1:])
return paths
def evaluate(self, sess, data_manager, id_to_tag):
"""
:param sess: session to run the model
:param data: list of data
:param id_to_tag: index to tag name
:return: evaluate result
"""
results = []
trans = self.trans.eval()
for batch in data_manager.iter_batch():
strings = batch[0]
tags = batch[-1]
lengths, scores = self.run_step(sess, False, batch)
batch_paths = self.decode(scores, lengths, trans)
for i in range(len(strings)):
result = []
string = strings[i][:lengths[i]]
gold = iobes_iob([id_to_tag[int(x)] for x in tags[i][:lengths[i]]])
pred = iobes_iob([id_to_tag[int(x)] for x in batch_paths[i][:lengths[i]]])
#gold = iob_iobes([id_to_tag[int(x)] for x in tags[i][:lengths[i]]])
#pred = iob_iobes([id_to_tag[int(x)] for x in batch_paths[i][:lengths[i]]])
for char, gold, pred in zip(string, gold, pred):
result.append(" ".join([char, gold, pred]))
results.append(result)
return results
def evaluate_line(self, sess, inputs, id_to_tag):
#trans条件随机场分类得出的矩阵
trans = self.trans.eval(session=sess)
#score是[句数，字数，label数]
lengths, scores = self.run_step(sess, False, inputs)
#viterbi_decode 由显式序列scores和状态转移阵trans，求隐藏序列的最大概率路径，也即最短路径
batch_paths = self.decode(scores, lengths, trans)
tags = [id_to_tag[idx] for idx in batch_paths[0]]
return result_to_json(inputs[0][0], tags)
