链接:https://pan.baidu.com/s/1ZiMzKulcsEt2xo_a2XK1nw
提取码:9umt
import pandas as pd
import fool
import re
import random
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
# -----------------------------------------------------
# 加载停用词词典
stopwords = {}
with open(r'stopword.txt', 'r', encoding='utf-8') as fr:
for word in fr:
stopwords[word.strip()] = 0
# -----------------------------------------------------
# 定义类
class clf_model:
"""
该类将所有模型训练、预测、数据预处理、意图识别的函数包括其中
"""
# 初始化模块
def __init__(self):
self.model = "" # 成员变量,用于存储模型
self.vectorizer = "" # 成员变量,用于存储tfidf统计值
# 训练模块
def train(self):
"""
训练结果存储在成员变量中,没有return
"""
# 从excel文件读取训练样本
d_train = pd.read_excel("data_train.xlsx")
# 对训练数据进行预处理
d_train.sentence_train = d_train.sentence_train.apply(self.fun_clean)
print("训练样本 = %d" % len(d_train))
# 利用sklearn中的函数进行tifidf训练
self.vectorizer = TfidfVectorizer(analyzer="word",
token_pattern=r"(?u)\b\w+\b") # 注意,这里自己指定token_pattern,否则sklearn会自动将一个字长度的单词过滤筛除
features = self.vectorizer.fit_transform(d_train.sentence_train)
print("训练样本特征表长度为 " + str(features.shape))
# 使用逻辑回归进行训练和预测
self.model = LogisticRegression(C=10)
self.model.fit(features, d_train.label)
# 预测模块(使用模型预测)
def predict_model(self, sentence):
# --------------
# 对样本中没有点特殊情况做特别判断
if sentence in ["好的", "需要", "是的", "要的", "好", "要", "是"]:
return 1, 0.8
# --------------
sent_features = self.vectorizer.transform([sentence])
pre_test = self.model.predict_proba(sent_features).tolist()[0]
clf_result = pre_test.index(max(pre_test))
score = max(pre_test)
return clf_result, score
# 预测模块(使用规则)
def predict_rule(self, sentence):
"""
如果模型训练出现异常,可以使用规则进行预测,同时也可以让学员融合"模型"及"规则"的预测方式
:param sentence:
:return 预测结果:
"""
sentence = sentence.replace(' ', '')
if re.findall(r'不需要|不要|停止|终止|退出|不买|不定|不订', sentence):
return 2, 0.8
elif re.findall(r'订|定|预定|买|购', sentence) or sentence in ["好的", "需要", "是的", "要的", "好", "要", "是"]:
return 1, 0.8
else:
return 0, 0.8
# 预处理函数
def fun_clean(self, sentence):
"""
预处理函数
:输入 用户输入语句:
:输出 预处理结果:
"""
# 使用foolnltk进行实体识别
words, ners = fool.analysis(sentence)
# 对识别结果按长度倒序排序
ners = ners[0].sort(key=lambda x: len(x[-1]), reverse=True)
# 如果有实体被识别出来,就将实体的字符串替换成实体类别的字符串(目的是看成一类单词,看成一种共同的特征)
if ners:
for ner in ners:
sentence = sentence.replace(ner[-1], ' ' + ner[2] + ' ')
# 分词,并去除停用词
word_lst = [w for w in fool.cut(sentence)[0] if w not in stopwords]
output_str = ' '.join(word_lst)
output_str = re.sub(r'\s+', ' ', output_str)
return output_str.strip()
# 分类主函数
def fun_clf(self, sentence):
"""
意图识别函数
:输入 用户输入语句:
:输出 意图类别,分数:
"""
# 对用户输入进行预处理
sentence = self.fun_clean(sentence)
# 得到意图分类结果(0为“查询”类别,1为“订票”类别,2为“终止服务”类别)
clf_result, score = self.predict_model(sentence) # 使用训练的模型进行意图预测
# clf_result, score = self.predict_rule(sentence) # 使用规则进行意图预测(可与用模型进行意图识别的方法二选一)
return clf_result, score
def fun_replace_num(sentence):
"""
替换时间中的数字(目的是便于实体识别包fool对实体的识别)
:param sentence:
:return sentence:
"""
# 定义要替换的数字
time_num = {"一": "1", "二": "2", "三": "3", "四": "4", "五": "5", "六": "6", "七": "7", "八": "8", "九": "9", "十": "10",
"十一": "11", "十二": "12"}
for k, v in time_num.items():
sentence = sentence.replace(k, v)
return sentence
def slot_fill(sentence, key=None):
"""
填槽函数(该函数从sentence中寻找需要的内容,完成填槽工作)
:param sentence:
:return slot: (填槽的结果)
"""
slot = {}
# 进行实体识别
words, ners = fool.analysis(sentence)
to_city_flag = 0 # flag为1代表找到到达城市(作用:当找到到达城市时,默认句子中另一个城市信息是出发城市)
for ner in ners[0]:
# 首先对time类别的实体进行信息抽取填槽工作
if ner[2] == 'time':
# --------------------
# 寻找日期的关键词
date_content = re.findall(
r'后天|明天|今天|大后天|周末|周一|周二|周三|周四|周五|周六|周日|本周一|本周二|本周三|本周四|本周五|本周六|本周日|下周一|下周二|下周三|下周四|下周五|下周六|下周日|这周一|这周二|这周三|这周四|这周五|这周六|这周日|\d{,2}月\d{,2}号|\d{,2}月\d{,2}日',
ner[-1])
slot["date"] = date_content[0] if date_content else ""
# 完成日期的填槽
# --------------------
# --------------------
# 寻找具体时间的关键词
time_content = re.findall(r'\d{,2}点\d{,2}分|\d{,2}点钟|\d{,2}点', ner[-1])
# 寻找上午下午的关键词
pmam_content = re.findall(r'上午|下午|早上|晚上|中午|早晨', ner[-1])
slot["time"] = pmam_content[0] if pmam_content else "" + time_content[0] if time_content else ""
# 完成时间的填槽
# --------------------
# 对location类别对实体进行信息抽取填槽工作
if ner[2] == 'location':
# --------------------
# 开始对城市填槽
# 如果没有指定槽位
if key is None:
if re.findall(r'(到|去|回|回去)%s' % (ner[-1]), sentence):
to_city_flag = 1
slot["to_city"] = ner[-1]
continue
if re.findall(r'从%s|%s出发' % (ner[-1], ner[-1]), sentence):
slot["from_city"] = ner[-1]
elif to_city_flag == 1:
slot["from_city"] = ner[-1]
# 如果指定了槽位
elif key in ["from_city", "to_city"]:
slot[key] = ner[-1]
# 完成出发城市、到达城市的填槽工作
# --------------------
return slot
def fun_wait(clf_obj):
"""
等待询问函数
:输入 None:
:输出 用户意图类别:
"""
# 等待用户输入
print("\n\n\n")
print("-------------------------------------------------------------")
print("-------------------------------------------------------------")
print("Starting ...")
sentence = input("客服:请问需要什么服务?(时间请用12小时制表示)\n")
# 对用户输入进行意图识别
clf_result, score = clf_obj.fun_clf(sentence)
return clf_result, score, sentence
def fun_search(clf_result, sentence):
"""
为用户查询余票
:param clf_result:
:param sentence:
:return: 是否有票
"""
# 定义槽存储空间
name = {"time": "出发时间", "date": "出发日期", "from_city": "出发城市", "to_city": "到达城市"}
slot = {"time": "", "date": "", "from_city": "", "to_city": ""}
# 使用用户第一句话进行填槽
sentence = fun_replace_num(sentence)
slot_init = slot_fill(sentence)
for key in slot_init.keys():
slot[key] = slot_init[key]
# 对未填充对槽位,向用户提问,进行针对性填槽
while "" in slot.values():
for key in slot.keys():
if slot[key] == "":
sentence = input("客服:请问%s是?\n" % (name[key]))
sentence = fun_replace_num(sentence)
slot_cur = slot_fill(sentence, key)
for key in slot_cur.keys():
if slot[key] == "":
slot[key] = slot_cur[key]
# 查询是否有票,并答复用户(本次查询是否有票使用随机数完成)
if random.random() > 0.5:
print("客服:%s%s从%s到%s的票充足" % (slot["date"], slot["time"], slot["from_city"], slot["to_city"]))
# 返回1表示有票
return 1
else:
print("客服:%s%s从%s到%s无票" % (slot["date"], slot["time"], slot["from_city"], slot["to_city"]))
print("End !!!")
print("-------------------------------------------------------------")
print("-------------------------------------------------------------")
# 返回0表示无票
return 0
def fun_book():
"""
为用户订票
"""
print("客服:已为您完成订票。\n\n\n")
print("End !!!")
print("-------------------------------------------------------------")
print("-------------------------------------------------------------")
if __name__ == "__main__":
# 实例化对象
clf_obj = clf_model()
clf_obj.train()
threshold = 0.55 # 用户定义阈值(当分类器分类的分数大于阈值才采纳本次意图分类结果,目的是排除分数过低的意图分类结果)
while 1:
clf_result, score, sentence = fun_wait(clf_obj)
# -------------------------------------------------------------------------------
# 状态转移条件(等待-->等待):用户输入未达到“查询”、“订票”类别的阈值 OR 被分类为“终止服务”
# -------------------------------------------------------------------------------
if score 查询):用户输入分类为“查询” OR “订票”
# -------------------------------------------------------------------------------
else:
search_result = fun_search(clf_result, sentence)
if search_result == 0:
continue
else:
# 等待用户输入
sentence = input("客服:需要为您订票吗?\n")
# 对用户输入进行意图识别
clf_result, score = clf_obj.fun_clf(sentence)
# -------------------------------------------------------------------------------
# 状态转移条件(查询-->订票):FUN_SEARCH返回有票 AND 用户输入分类为“订票”
# -------------------------------------------------------------------------------
if clf_result == 1:
fun_book()
continue
运行结果
作者:GlassySky0816