def handle(self):
# 初始化语义服务器
# 从 qa 初始化
# 从 qa_sql 初始化
# robot = Robot(path=getConfig("path", "db"), password=None)
while True:
# self.request is the TCP socket connected to the client
self.data = self.request.recv(2048)
if not self.data:
break
print("\n{} wrote:".format(self.client_address[0]))
self.data = self.data.decode("UTF-8")
print("Data:\n", self.data)
# step 1.Bytes to json obj and extract question
json_data = json.loads(self.data)
# step 2.Get answer
if "ask_content" in json_data.keys():
answer = self.robot.search(question=json_data["ask_content"],
userid=json_data["userid"],
key=json_data["key"])
info = json_data["ask_content"]
# 其中 result['picurl'] 为 xml 格式
result = answer2xml(answer)
elif "config_content" in json_data.keys():
answer = self.robot.configure(info=json_data["config_content"],
userid=json_data["userid"],
key=json_data["key"])
info = json_data["config_content"]
result = answer
print(answer)
print(result)
# step 3.Send
try:
self.request.sendall(json.dumps(result).encode("UTF-8"))
except:
with open(logpath, "a", encoding="UTF-8") as file:
file.write(
get_current_time("%Y-%m-%d %H:%M:%S") + "\n" +
"发送失败\n")
# 追加日志
with open(logpath, "a", encoding="UTF-8") as file:
# 写入接收数据中的内容字段
file.write(
get_current_time("%Y-%m-%d %H:%M:%S") + "\n" + info + "\n")
# 写入正常问答
if "ask_content" in json_data.keys():
for key in [
"question", "content", "behavior", "url",
"context", "parameter", "picurl"
]:
file.write(key + ": " + str(result[key]) + "\n")
# 写入配置信息
elif "config_content" in json_data.keys():
file.write("Config: " + json.dumps(result) + "\n")
file.write("\n")
def __init__(self, password="******", userid="A0001"):
self.graph = Graph(getConfig("neo4j", "uri"), password=password)
print(self.graph)
self.selector = NodeSelector(self.graph)
# self.locations = get_navigation_location()
self.is_scene = False
self.user = None
# self.user = self.selector.select("User", userid=userid).first()
# self.usertopics = self.get_usertopics(userid=userid)
# self.address = get_location_by_ip(self.user['city'])
self.topic = ""
self.behavior = 0 # 场景类型 Add in 2018-6-7
self.last_step_error = False # 在场景内上一个问答是否正常 Add in 2018-6-12
self.qa_id = get_current_time()
self.qmemory = deque(maxlen=10)
self.amemory = deque(maxlen=10)
self.pmemory = deque(maxlen=10)
# TODO：判断意图是否在当前话题领域内
self.change_attention = ["换个话题吧", "太无聊了", "没意思", "别说了"]
self.cmd_end_scene = ["退出业务场景", "退出场景", "退出", "返回", "结束", "发挥"]
self.cmd_previous_step = ["上一步", "上一部", "上一页", "上一个"]
self.cmd_next_step = ["下一步", "下一部", "下一页", "下一个"]
self.cmd_repeat = ["重复", "再来一个", "再来一遍", "你刚说什么", "再说一遍", "重来"]
self.yes = ["是", "是的", "对", "对的", "好的", "YES", "yes", "结束了"]
self.no = ["没", "没有", "否", "不", "没有结束", "没结束"]
self.do_not_know = [
"这个问题太难了，{robotname}还在学习中", "这个问题{robotname}不会，要么我去问下",
"您刚才说的是什么，可以再重复一遍吗", "{robotname}刚才走神了，一不小心没听清",
"{robotname}理解的不是很清楚啦，你就换种方式表达呗", "不如我们换个话题吧", "咱们聊点别的吧",
"{robotname}正在学习中", "{robotname}正在学习哦", "不好意思请问您可以再说一次吗",
"额，这个问题嘛。。。", "{robotname}得好好想一想呢", "请问您说什么", "您问的问题好有深度呀",
"{robotname}没有听明白，您能再说一遍吗"
]
def add_to_memory(Q="Q",
words=None,
tags=None,
content=None,
username="******"):
"""
将用户当前语句的语义分析结果加入信息记忆
"""
# 添加到记忆链，每一类型的节点都需要重构
qaID = username + "_" + get_current_time()
for semantic_tree in content:
generate_tree(NodeClass="Memory",
Q=Q,
words=words,
content=semantic_tree,
tags=tags,
username=username)
def test_txt():
current_user = "******"
print("QA测试......")
filename = "log/QA_" + get_current_time() + ".md"
f = open(filename, "w")
f.write("标签：测试文档\n#QA测试：\n>Enter the QA mode...\n")
while True:
try:
sentence = input("\n>>")
# 基于语义模式匹配
answer = search_database(question=sentence, username=current_user)
# 基于上下文理解
# answer = understand_context(question=sentence, username=current_user)
print("R: " + answer)
f.write("`>>" + sentence + "`\n")
f.write("`" + "A: " + answer + "`\n")
except KeyboardInterrupt:
f.close()
def extract_synonym(question, subgraph, pattern="wf"):
"""
从图形数据库返回的搜索结果列表选取相似度最高的语义Jaccard匹配同义句
pattern可选'w'-分词, 't'-关键词, 'wf'-分词标签, 'tf-关键词标签'
"""
similarity = []
answer = "您好！请问有什么可以帮您的吗？"
do_not_know = [
"正在学习中", "小民正在学习哦", "不好意思请问您可以再说一次吗", "额，这个问题嘛。。。", "我得好好想一想呢",
"请问您说什么", "。。。", "您问的问题好有深度呀"
]
command = {
"time": ["几点了", "现在几点了", "现在时间", "现在时间是多少"],
"cmd": ["打电话", "打电话给"]
}
if question in command["time"]:
answer = get_current_time("现在是%Y年%m月%d日%H点%M分%S秒")
return answer
elif question in command["cmd"]:
answer = "正在连接中，请稍候"
return answer
sv1 = synonym_cut(question, pattern)
for node in subgraph:
sv2 = synonym_cut(node["Q"], pattern)
ss = semantic_similarity(sv1, sv2, pattern="sj")
similarity.append(ss)
max_similarity = max(similarity)
print("Similarity Score: " + str(max_similarity))
index = similarity.index(max_similarity)
Q = subgraph[index]["Q"]
A = subgraph[index]["A"]
print("Q: " + Q)
if max_similarity > 0.2:
if isinstance(A, list):
answer = random_item(A)
else:
# 随机回答
answer = random_item(do_not_know)
return answer
def add_to_memory(self, question="question", userid="A0001"):
"""Add user question to memory.
将用户当前对话加入信息记忆。
Args:
question: 用户问题。
Defaults to "question".
userid: 用户唯一标识。
Defaults to "userid".
"""
previous_node = self.graph.find_one("Memory", "qa_id", self.qa_id)
self.qa_id = get_current_time()
node = Node("Memory",
question=question,
userid=userid,
qa_id=self.qa_id)
if previous_node:
relation = Relationship(previous_node, "next", node)
self.graph.create(relation)
else:
self.graph.create(node)
# def extract_navigation(self, question):
"""Extract navigation from question。从问题中抽取导航地点。
"""
graph = Graph("http://localhost:7474/db/data/", password="******")
sv = []
for word in words:
word_node = graph.find_one("Synonym", "word", word)
if word_node:
tag_node = graph.find_one("SynonymTag", "name", word_node["tag"])
tag_words = tag_node["words"]
synonym_random = random_item(tag_words)
sv.append(synonym_random)
return sv
if __name__ == '__main__':
print("语义相似度测试......")
filename = "log/SemanticSimilarity_" + get_current_time() + ".md"
f = open(filename, "w")
f.write("标签：测试文档\n#语义相似度测试：\n>Enter the SemanticSimilarity mode...\n")
while True:
try:
sentence1 = input("\nsentence1\n>>")
sentence2 = input("sentence2\n>>")
w1 = synonym_cut(sentence1, 'w')
w2 = synonym_cut(sentence2, 'w')
sv1 = synonym_cut(sentence1, 'wf')
sv2 = synonym_cut(sentence2, 'wf')
print(w1, w2)
similarity = semantic_similarity(w1, w2)
print("similarity: " + str(similarity))
print(sv1, sv2)
score = word_similarity(w1, w2)
sv_rows.append(score)
sv_matrix.append(sv_rows)
sv_rows = []
matrix = numpy.mat(sv_matrix)
result = sum_cosine(matrix, 0.6)
total = result["total"]
total_dif = result["total_dif"]
m = result["m"]
similarity = total/(total + m*(1-total_dif))
return similarity
if __name__ == '__main__':
print("向量语义相似度测试......")
filename = "log/VecSimilarity_" + get_current_time() + ".md"
f = open(filename, "w")
f.write("标签：测试文档\n#向量语义相似度测试：\n>Enter the VecSimilarity mode...\n")
while True:
try:
sentence1 = input("\nsentence1\n>>")
sentence2 = input("sentence2\n>>")
similarity = vec_jaccard(sentence1, sentence2, 'w')
print("similarity: " + str(similarity))
similarity = vec_jaccard(sentence1, sentence2, 't')
print("similarity: " + str(similarity))
f.write("`>>" + sentence1 + "`\n")
f.write("`>>" + sentence2 + "`\n")
f.write("`" + "similarity: " + str(similarity) + "`\n")