fix the identity recognition issue for minority population
This commit is contained in:
weiweiw
parent
1799c04a92
commit
4e920f5542
|
|
@ -32,40 +32,116 @@ class IdentityCardExtractor(Extractor):
|
|||
logger.error(e)
|
||||
return {}
|
||||
|
||||
# def extract_textbyPaddle(self, text:str)->dict:
|
||||
# try:
|
||||
# patterns = {
|
||||
# "issuingAuthority": r"签发机关\n*(.+?)\n",
|
||||
# "validTime": r"有效期限\n*(.+?)\n",
|
||||
# "name": r"姓名(.*?)\n", #####
|
||||
# "gender": r"(\S)民族",
|
||||
# "ethnicity": r"民族(\S+)",
|
||||
# "dateOfBirth": r"(\d+年\d+月\d+日)",
|
||||
# "address": r"(住址|佳址)\s*(.*?)公民身份号码",
|
||||
# "idNumber": r"(\d{18}|\d{17}[Xx])"
|
||||
# }
|
||||
#
|
||||
# tempText = self.remove_blank_lines(text)
|
||||
# # 提取信息
|
||||
# info = {}
|
||||
# for key, pattern in patterns.items():
|
||||
# match = re.search(pattern, tempText,re.DOTALL)
|
||||
# if match:
|
||||
# if "address" == key:
|
||||
# tempStr = match.group(2).strip()
|
||||
# else:
|
||||
# tempStr = match.group(1).strip()
|
||||
# info[key] = tempStr.replace("\n", "")
|
||||
# return info
|
||||
# except Exception as e:
|
||||
# print(e)
|
||||
# logger.error(e)
|
||||
# return {}
|
||||
|
||||
def extract_textbyPaddle(self, text:str)->dict:
|
||||
try:
|
||||
patterns = {
|
||||
"issuingAuthority": r"签发机关\n*(.+?)\n",
|
||||
"validTime": r"有效期限\n*(.+?)\n",
|
||||
"name": r"姓名(.*?)\n", #####
|
||||
"gender": r"(\S)民族",
|
||||
"ethnicity": r"民族(\S+)",
|
||||
"dateOfBirth": r"(\d+年\d+月\d+日)",
|
||||
"address": r"(住址|佳址)\s*(.*?)公民身份号码",
|
||||
"idNumber": r"(\d{18}|\d{17}[Xx])"
|
||||
}
|
||||
result = {}
|
||||
# 提取签发机关
|
||||
issuing_authority = re.search(r"签发机关\n*(.+?)\n", text, re.DOTALL)
|
||||
if issuing_authority:
|
||||
result["issuingAuthority"] = issuing_authority.group(1).strip()
|
||||
|
||||
tempText = self.remove_blank_lines(text)
|
||||
# 提取信息
|
||||
info = {}
|
||||
for key, pattern in patterns.items():
|
||||
match = re.search(pattern, tempText,re.DOTALL)
|
||||
if match:
|
||||
if "address" == key:
|
||||
tempStr = match.group(2).strip()
|
||||
# 提取有效期限
|
||||
valid_time = re.search(r"有效期限\n*(\d{4}\.\d{2}\.\d{2}-\S+)", text, re.DOTALL)
|
||||
if valid_time:
|
||||
result["validTime"] = valid_time.group(1).strip()
|
||||
|
||||
# 提取姓名
|
||||
name = re.search(r"姓名\s*(.*?)\n", text,re.DOTALL)
|
||||
if name:
|
||||
tempName = name.group(1).strip()
|
||||
if tempName in "性别男" or tempName in "性别女":
|
||||
name = re.search(r"(.*?)\s*姓名", text, re.DOTALL)
|
||||
result["name"] = name.group(1).strip()
|
||||
else:
|
||||
result["name"] = name.group(1).strip()
|
||||
|
||||
else:
|
||||
name = re.search(r"米名(\S*)址\s*(\S+)", text, re.DOTALL)
|
||||
if name:
|
||||
result["name"] = name.group(2).strip()
|
||||
else:
|
||||
name = re.search(r"名\s*(\S+)\s*姓\s*", text, re.DOTALL)
|
||||
if name:
|
||||
result["name"] = name.group(1).strip()
|
||||
else:
|
||||
tempStr = match.group(1).strip()
|
||||
info[key] = tempStr.replace("\n", "")
|
||||
return info
|
||||
name = re.search(r"(\S+)\s*(男|女|性别)", text, re.DOTALL)
|
||||
if name:
|
||||
result["name"] = name.group(1).strip()
|
||||
|
||||
# 提取民族
|
||||
ethnicity = re.search(r"民\s*族\s*(\S+)", text, re.DOTALL)
|
||||
if ethnicity:
|
||||
result["ethnicity"] = ethnicity.group(1).strip()
|
||||
|
||||
# 提取地址
|
||||
address = re.search(r"(住址|佳址)(.*?)公民身份号码", text, re.DOTALL)
|
||||
if address:
|
||||
result["address"] = address.group(2).strip().replace("\n", "")
|
||||
else:
|
||||
address = re.search(r"(\S+省)(.*?)公民身份号码", text, re.DOTALL)
|
||||
if address:
|
||||
result["address"] = address.group(1).strip().replace("\n", "") + address.group(2).strip().replace("\n", "")
|
||||
else:
|
||||
address = re.search(r"(\S+市)(.*?)公民身份号码", text, re.DOTALL)
|
||||
if address:
|
||||
result["address"] = address.group(1).strip().replace("\n", "") + address.group(2).strip().replace("\n", "")
|
||||
else:
|
||||
address = re.search(r"(\S+县)(.*?)公民身份号码", text, re.DOTALL)
|
||||
if address:
|
||||
result["address"] = address.group(1).strip().replace("\n", "") + address.group(2).strip().replace("\n", "")
|
||||
if result["address"]:
|
||||
result["address"] = re.sub(r'[A-Z]', '', result["address"])
|
||||
|
||||
# 提取身份证号码
|
||||
id_number = re.search(r"(\d{18}|\d{17}[Xx])", text, re.DOTALL)
|
||||
if id_number:
|
||||
result["idNumber"] = id_number.group(1).strip()
|
||||
|
||||
if result["idNumber"]:
|
||||
# 提取出生日期
|
||||
result["dateOfBirth"] = self.extract_birthday_from_id(result["idNumber"])
|
||||
# 提取性别
|
||||
result["gender"] = self.get_gender_from_id(result["idNumber"])
|
||||
|
||||
return result
|
||||
except Exception as e:
|
||||
print(e)
|
||||
logger.error(e)
|
||||
return {}
|
||||
|
||||
|
||||
class InvoiceExtractor(Extractor):
|
||||
def extract_text(self,text:str)->dict:
|
||||
pass
|
||||
class InvoiceExtractor(Extractor):
|
||||
def extract_text(self,text:str)->dict:
|
||||
pass
|
||||
|
||||
|
||||
# text = """中华人民共和国
|
||||
|
|
@ -96,7 +172,145 @@ class InvoiceExtractor(Extractor):
|
|||
# 公民身份号码
|
||||
# 34220119941017327X
|
||||
# """
|
||||
# extractor = IdentityCardExtractor()
|
||||
|
||||
# text = """中华人民共和国
|
||||
# 居民身份证
|
||||
# 签发机关
|
||||
# 盐源县公安局
|
||||
# 出
|
||||
# 有效期限
|
||||
# 2013.06.18-2033.06.18
|
||||
# 坐姓5性中出住
|
||||
# 米名年别到生委址
|
||||
# 江六斤
|
||||
# 男民族彝
|
||||
# 半口
|
||||
# 1980
|
||||
# 年4月20日
|
||||
# 四川省盐源县盖租乡阿石
|
||||
# 村6组4号
|
||||
# 公民身份号码
|
||||
# 513423198004203995
|
||||
# """
|
||||
#
|
||||
# text = """姓名韩邀宇
|
||||
# 性别男
|
||||
# 民族汉
|
||||
# 出生
|
||||
# 1999年12月15日
|
||||
# 住址
|
||||
# 安徽省太和县税镇镇十里
|
||||
# 沟村委会中韩村16号
|
||||
# 公民身份号码
|
||||
# 34122219991215183X"""
|
||||
#
|
||||
#
|
||||
# text = """中华人民共和国
|
||||
# 居民身份证
|
||||
# 签发机关
|
||||
# 木里县公安局
|
||||
# 有效期限
|
||||
# 2016.12.05-2026.12.05
|
||||
# 姓性出住
|
||||
# 马加加
|
||||
# 女
|
||||
# 民族彝
|
||||
# 1998年1月2日
|
||||
# 址
|
||||
# 四川省木里藏族自治县博
|
||||
# 科乡八科村麻窝地组18号
|
||||
# 公民身份号码
|
||||
# 513422199801023821"""
|
||||
|
||||
# text = """
|
||||
# 毛阿卡
|
||||
# 姓名
|
||||
# 性别男
|
||||
# 民族彝
|
||||
# 3
|
||||
# 出生
|
||||
# 1976年3月3日
|
||||
# 住址
|
||||
# 四川省木里藏族自治县耗
|
||||
# 牛坪乡泥珠村下泥珠组55
|
||||
# 号
|
||||
# 中国LHINA
|
||||
# 公民身份号码
|
||||
# 513423197603033997
|
||||
# 中华人民共和国
|
||||
# 居民身份证
|
||||
# 签发机关
|
||||
# 木里县公安局
|
||||
# 有效期限
|
||||
# 2017.06.06-2037.06.06
|
||||
# """
|
||||
# text = """中华人民共和国
|
||||
# 居民身份证
|
||||
# 国
|
||||
# 签发机关
|
||||
# 西昌市公安局
|
||||
# 有效期限
|
||||
# 2008.09.01-2028.09.01
|
||||
# 逆姓5性中出日住
|
||||
# 米名丰别刺生更址
|
||||
# 祝九根惹
|
||||
# 心
|
||||
# 民
|
||||
# 族彝
|
||||
# 男
|
||||
# E
|
||||
# 1978
|
||||
# 四川省西昌市磨盘乡大厂
|
||||
# 村4组27号
|
||||
# 公民身份号码
|
||||
# 513401197807087411
|
||||
# """
|
||||
# text = """中华人民共和国
|
||||
# 居民身份证
|
||||
# 签发机关
|
||||
# 木里县公安局
|
||||
# 有效期限
|
||||
# 2020.03.16-2025.03.16
|
||||
# 名
|
||||
# 蒋子古
|
||||
# 姓
|
||||
# 男
|
||||
# 民族彝
|
||||
# 出生
|
||||
# 2005年1月4日
|
||||
# 住址
|
||||
# 四川省木里藏族自治县耗
|
||||
# 牛坪乡泥珠村5组29号
|
||||
# 公民身份号码
|
||||
# 513422200501044415
|
||||
# """
|
||||
# extractor = IdentityCardExtractor()
|
||||
# jsonstring = extractor.extract_textbyPaddle(text)
|
||||
# print(jsonstring)
|
||||
|
||||
|
||||
|
||||
|
||||
# text = """
|
||||
# 姓名张三
|
||||
# 性别男
|
||||
# 民族汉
|
||||
# # """
|
||||
# text = """
|
||||
# 江六斤
|
||||
# 女民族彝
|
||||
# """
|
||||
# # 尝试使用 r"性别(\S)" 匹配性别
|
||||
# gender_match = re.search(r'性别(\S)|(\S)民族', text)
|
||||
# if gender_match:
|
||||
# gender = gender_match.group(0)
|
||||
# print("性别是:", gender)
|
||||
# # else:
|
||||
# # # 如果匹配不到，尝试使用 r"(\S)民族" 匹配
|
||||
# # gender_match = re.search(r'(\S)民族', text)
|
||||
# # if gender_match:
|
||||
# # gender = gender_match.group(1)
|
||||
# # print("性别是:", gender)
|
||||
# else:
|
||||
# print("未找到性别信息")
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue