def get_process_data(self, file):
test_data = ChangeDataType.json_to_dict(rootPath + "\\testdata\\uidata\\" + file)
# print(test_data)
# # 遍历item_process.json文件中的项目与项目问诊内容
# for key in test_data:
# print(key + ':' + str(test_data[key]))
return test_data
def get_eye_intent(self, api_url, target_file, test_data_file,
result_file):
"""
通过抽取测试集的数据，调用意图接口，得出的测试结果，在调用函数获取每个target的准确率，召回率，F1
:param target_file: 储存target的文件
:param data_file: 储存接口结果数据的文件
"""
# 获取测试集的data
test_data = ChangeDataType.csv_to_dict(rootPath +
"\\testdata\\apidata\\" +
test_data_file)
score_list = []
re_intent_list = []
bz_intent_list = []
tf_list = []
re_intent = ""
tf = ""
# 循环读取sentence，intent
for idx, temp in test_data.iterrows():
intent = temp["intention"]
sentence = temp["sentence"]
# 发起请求
url = api_url.format(sentence)
try:
r = requests.get(url, timeout=50)
result = r.json()
re_intent = result["data"]["intent"] # 获取返回data的intent
print(re_intent)
score = result["data"]["intent_probability"] # 获取返回data的score
tf = CommonFunction.get_tf(intent, re_intent)
except Exception as e:
score = "bad request"
print(e)
# self.logging.info("句子：" + sentence + "---预期意图：" + intent
# + "---实际意图：" + re_intent + "---是否一致：" + tf)
# 拼接结果数据
score_list.append(score)
bz_intent_list.append(intent)
re_intent_list.append(re_intent)
tf_list.append(tf)
test_data["re_intent"] = re_intent_list
test_data["score"] = score_list
# 调用方法，拼接test_data值
test_data = CommonFunction.get_collections(test_data, tf_list)
now = time.strftime('%y_%m_%d-%H_%M_%S')
# 输出excel
test_data.to_excel(rootPath + '\\testresults\\resultfile\\' + now +
result_file,
index=False,
encoding="utf-8")
GetIntent.get_intent_result(self, target_file, bz_intent_list,
re_intent_list)
def get_pro_intent(self, api_url, target_file, test_data_file,
result_file):
"""
通过抽取测试集的数据，调用意图接口，得出的测试结果，在调用函数获取每个target的准确率，召回率，F1
:param target_file: 储存target的文件
:param data_file: 储存接口结果数据的文件
"""
# 获取测试集的data
test_data = ChangeDataType.csv_to_dict(
rootPath + "\\testdata\\apidata\\intent\\" + test_data_file)
score_list = []
re_intent_list = []
exp_intent_list = []
tf_list = []
# 循环读取sentence，intent
for idx, temp in test_data.iterrows():
intent = temp["intent"]
sentence = temp["sentence"]
headers = {
'Authorization':
"eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJjb21wYW55X2lkIjoxLCJyb2JvdF9pZCI6MSwiZXhwIjoxNTg5MzMyNTIzfQ.Durc3V9XA99BejXc2ZOzspPU-JJCY1nUUjceICwBWNE",
'Content-Type': 'application/x-www-form-urlencoded'
}
data = {
'dialog': sentence,
'client_id': "zltest",
}
url = api_url.format(sentence) # 接口请求
try:
r = requests.post(url, data=data, headers=headers, timeout=50)
result = r.json()
re_intent = result["ner_result"]["ner"]["intent"]["Value"][0][
"Value"] # 获取返回data的intent
print(re_intent)
tf = CommonFunction.get_tf(intent, re_intent)
except Exception as e:
score = "bad request"
exp_intent_list.append(intent)
re_intent_list.append(re_intent)
tf_list.append(tf)
test_data["re_intent"] = re_intent_list
# 调用方法，拼接test_data值
test_data = CommonFunction.get_collections(test_data, tf_list)
now = time.strftime('%y_%m_%d-%H_%M_%S')
# 输出excel
test_data.to_excel(rootPath + '\\testresults\\resultfile\\' + now +
result_file,
index=False,
encoding="utf-8")
GetIntent.get_intent_result(self, target_file, exp_intent_list,
re_intent_list)
def get_pro_symptom(self, api_url, test_data_file, result_file):
test_data = ChangeDataType.json_to_dict(rootPath + "\\testdata\\apidata\\" + test_data_file)
key_list = list(test_data.keys())
value_list = list(test_data.values())
result_value_list = []
tf_list = []
tf = ""
for key, value in test_data.items():
url = api_url.format(key)
try:
r = requests.get(url, timeout=50)
result = r.json()
result_value = result["norm_symptoms"]
tf = CommonFunction.get_tf(result_value, value[0])
except Exception as e:
print(e)
# self.logging.info("口语症状：" + key + "---预期标准症状：" + value[0] + "---实际标准症状："
# + result_value + "---是否一致：" + tf)
result_value_list.append(result_value)
tf_list.append(tf)
# Binary.binary_plot_curve(value_list, result_value_list)
now = time.strftime('%y_%m_%d-%H_%M_%S')
workbook = xlwt.Workbook()
sheet1 = workbook.add_sheet('sheet1', cell_overwrite_ok=True)
sheet1.write(0, 0, "口语症状")
sheet1.write(0, 1, "预期标准症状")
sheet1.write(0, 2, "实际标准症状")
sheet1.write(0, 3, "是否一致")
sheet1.write(0, 4, "总数")
sheet1.write(1, 4, len(key_list))
sheet1.write(0, 5, "一致数")
sheet1.write(1, 5, tf_list.count("TRUE"))
sheet1.write(0, 6, "不一致数")
sheet1.write(1, 6, tf_list.count("FALSE"))
sheet1.write(0, 7, "一致率")
sheet1.write(1, 7, "{:.2f}%".format(tf_list.count("TRUE") / len(key_list)) * 100)
sheet1.write(0, 8, "不一致率")
sheet1.write(1, 8, "{:.2f}%".format(tf_list.count("FALSE") / len(key_list)) * 100)
print("总数：", len(tf_list), " 一致数：", tf_list.count("TRUE"), " 不一致数：", tf_list.count("FALSE"), " 一致率：",
"{:.2f}%".format(tf_list.count("TRUE") / len(tf_list) * 100), " 不一致率：",
"{:.2f}%".format(tf_list.count("FALSE") / len(tf_list) * 100))
for i in range(0, len(key_list)):
sheet1.write(i + 1, 0, key_list[i])
sheet1.write(i + 1, 1, value_list[i])
sheet1.write(i + 1, 2, result_value_list[i])
sheet1.write(i + 1, 3, tf_list[i])
workbook.save(rootPath + '\\testresults\\resultfile\\' + now + result_file)
def get_sentence_similarity(self, test_data_file, result_file,
compare_score_file):
# self.logging = Logging()
test_data = ChangeDataType.csv_to_dict(rootPath +
"\\testdata\\apidata\\" +
test_data_file)
score_list = []
re_score_list = []
tf_list = []
re_score = ""
tf = ""
lb_list = []
for idx, temp in test_data.iterrows():
label = int(temp["label"])
str1 = temp["症状a"]
str2 = temp["症状b"]
url = "http://192.168.1.74:8233/bert_similarity/v2?str1={}&str2={}&model".format(
str1, str2)
try:
r = requests.get(url, timeout=50)
result = r.json()
score = result["data"]["score"]
compare_score_list = CommonFunction.get_target(
(self, compare_score_file))
for i in range(0, len(compare_score_list)):
re_score = CommonFunction.get_re_score(
score, float(compare_score_list[i]))
tf = CommonFunction.get_tf(re_score, label)
except Exception as e:
score = "bad request"
print(score)
# self.logging.info("症状1：" + str1 + "---症状2：" + str2 + "---预期分数："
# + str(label) + "---实际分数：" + str(re_score) + "---是否一致：" + tf)
score_list.append(score)
re_score_list.append(re_score)
tf_list.append(tf)
lb_list.append(label)
test_data["score"] = score_list
test_data["re_score"] = re_score_list
test_data = CommonFunction.get_collections(test_data, tf_list)
Binary.binary_plot_curve(lb_list, re_score_list)
now = time.strftime('%y_%m_%d-%H_%M_%S')
test_data.to_excel(rootPath + '\\testresults\\resultfile\\' + now +
result_file,
index=False,
encoding="utf-8")
def get_answer(self, file):
"""
查看answer中的有效数据，sentence，ner_title,ner_context并做储备，后期做调用
:param file:answer文件
"""
sentence_list = []
each_ner_content = []
each_ner_title = []
each_ner_t_list = []
each_ner_c_list = []
test_data = ChangeDataType.csv_to_dict(rootPath +
"\\testdata\\pregnant\\" + file)
for idx, temp in test_data.iterrows():
# 获取句子，实体，实体类别，意图等
sentence = temp["句子"]
ner = temp["实体"]
ner_type = temp["实体类别"]
# 获取每个句子所有的实体类别：如果一个句子有多个实体类别，中间用$分割
ner_type_l = ner_type.split("$")
# 判断是否有多个实体类别
if len(ner_type_l) >= 2:
# 根据$拆分每个ner
ner_l = ner.split("$")
# 循环遍历拆分每个ner和内容
for i in range(1, len(ner_l)):
each_ner = ner_l[i].split(":")
# 拼接 每个ner的类别和内容
if i == 1:
each_ner_title.append(str(each_ner[0]))
each_ner_content.append(str(each_ner[1]))
if i >= 2:
each_ner_title.append("," + str(each_ner[0]))
each_ner_content.append("," + str(each_ner[1]))
# 拼接sentence，each_ner_content，each_ner_title
sentence_list.append(sentence)
each_ner_c_list.append(each_ner_content)
each_ner_t_list.append(each_ner_title)
# 只有单个实体
else:
# 拆分ner和内容
each_ner = ner.split("$")[1].split(":")
# 拼接 ner的类别和内容
each_ner_title.append(str(each_ner[0]))
each_ner_content.append(str(each_ner[1]))
# 拼接sentence，each_ner_content，each_ner_title
sentence_list.append(sentence)
each_ner_c_list.append(each_ner_content)
each_ner_t_list.append(each_ner_title)
# 重置 each_ner_content，each_ner_title
each_ner_content = []
each_ner_title = []
print(sentence_list)
print(each_ner_c_list)
print(each_ner_t_list)
# 建立表格，先将这三个list结果记录文档；sentence_list，each_ner_c_list，each_ner_t_list
workbook = xlwt.Workbook()
# 设置sheet名字
sheet1 = workbook.add_sheet('answer统计结果', cell_overwrite_ok=True)
# 设置三列title
sheet1.write(0, 0, "sentence")
sheet1.write(0, 1, "ner")
sheet1.write(0, 2, "ner_context")
# 循环填写三列内容
for i in range(0, len(sentence_list)):
sheet1.write(i + 1, 0, sentence_list[i])
sheet1.write(i + 1, 1, each_ner_c_list[i])
sheet1.write(i + 1, 2, each_ner_t_list[i])
# 保存表格
workbook.save(rootPath + '\\testresults\\resultfile\\' +
"answer_collection_result.xls")
return sentence_list, each_ner_c_list, each_ner_t_list
def get_question(self, file):
"""
查看question中的有效数据，sentence，ner_title,ner_context,intent，并做储备，后期做调用
:param file:question文件
"""
sentence_list = []
intent_list = []
each_intent_list = []
each_ner_content = []
each_ner_title = []
each_ner_t_list = []
each_ner_c_list = []
test_data = ChangeDataType.csv_to_dict(rootPath +
"\\testdata\\pregnant\\" + file)
for idx, temp in test_data.iterrows():
# 获取句子，实体，实体类别，意图等
sentence = temp["句子"]
ner = temp["实体"]
ner_type = temp["实体类别"]
intent = temp["意图"]
# 获取每个句子所有的实体类别：如果一个句子有多个实体类别，中间用$分割
ner_type_l = ner_type.split("$")
# 判断intent在target文件中，并且不等于无
if intent in GetRelation.get_intent_target(
self, "intent_target.txt") and ner_type != "无":
# 判断是否有多个实体类别
if len(ner_type_l) >= 2:
sentence_list.append(sentence)
# 拼接意图
each_intent_list.append(intent)
# 根据$拆分每个ner
ner_l = ner.split("$")
# 循环遍历拆分每个ner和内容
for i in range(1, len(ner_l)):
each_ner = ner_l[i].split(":")
# 拼接 每个ner的类别和内容
if i == 1:
each_ner_title.append(str(each_ner[0]))
each_ner_content.append(str(each_ner[1]))
if i >= 2:
each_ner_title.append("," + str(each_ner[0]))
each_ner_content.append("," + str(each_ner[1]))
intent_list.append(each_intent_list)
each_ner_c_list.append(each_ner_content)
each_ner_t_list.append(each_ner_title)
# 只有单个实体
else:
# ner_l = ner.split("$")
sentence_list.append(sentence)
# 拼接意图
each_intent_list.append(intent)
# 拆分ner和内容
each_ner = ner.split("$")[1].split(":")
# 拼接 ner的类别和内容
each_ner_title.append(str(each_ner[0]))
each_ner_content.append(str(each_ner[1]))
intent_list.append(each_intent_list)
each_ner_c_list.append(each_ner_content)
each_ner_t_list.append(each_ner_title)
# 拼接sentence，each_intent_list，each_ner_content，each_ner_title
# 重置each_ner_content，each_ner_title，each_intent_list
each_ner_content = []
each_ner_title = []
each_intent_list = []
print(sentence_list)
print(each_ner_c_list)
print(each_ner_t_list)
print(intent_list)
# 建立表格，先将这四个list结果记录文档；sentence_list，each_ner_c_list，each_ner_t_list，intent_list
workbook = xlwt.Workbook()
# 设置sheet名字
sheet1 = workbook.add_sheet('question统计结果', cell_overwrite_ok=True)
# 设置四列title
sheet1.write(0, 0, "sentence")
sheet1.write(0, 1, "ner")
sheet1.write(0, 2, "ner_context")
sheet1.write(0, 3, "intent")
# 循环填写四列内容
for i in range(0, len(sentence_list)):
sheet1.write(i + 1, 0, sentence_list[i])
sheet1.write(i + 1, 1, each_ner_c_list[i])
sheet1.write(i + 1, 2, each_ner_t_list[i])
sheet1.write(i + 1, 3, intent_list[i])
# 保存表格
workbook.save(rootPath + '\\testresults\\resultfile\\' +
"question_collection_result.xls")
return sentence_list, each_ner_c_list, each_ner_t_list, intent_list