class BrandRegTool(object):
def __init__(self, standard_brand_file, del_brand_file=None, exchange_brand_file=None, rule_brand_file=None):
if not os.path.exists(standard_brand_file):
raise Exception("%s does not exists!" % standard_brand_file)
try:
self.brand_loading_obj = BrandInfoLoading(standard_brand_file,del_brand_file, exchange_brand_file)
#直接读取进行过品牌扩展的文件
self.brand_idx_dict, self.idx_ori_brand_dict, self.name_ori_brand_dict, self.brand_cat1_dict, \
self.cat1_brand_dict, self.cat1_clean_brand_dict, self.brand_gmv_dict, \
self.cat1_dict = self.brand_loading_obj.brand_info_loading()
self.brand_rule_obj = BrandRefRuleOpt(rule_brand_file, self.idx_ori_brand_dict)
except Exception as e:
raise e
def product_name_del_word(self, ori_s_name):
del_word_dict = {"网红":'', '美的布谷':'', '珂丝兰珂润万物丝兰完美':'', '淘米家用':'', \
'家用':'', 'OLOMLB':'', 'TESIRIS':'', '文和友演出门票':'', '淘米':'', \
'小米推荐':'', '小米众筹':'', '小米家人':'', '内蒙':'', '牛肉干':'', \
'小米诺':'', '小米琦':'', '小米粒':'', '小米奇':'','小米生态链':'', \
'小米有品同款':'', '小米同款':''}
s_name = ori_s_name
for k, v in del_word_dict.items():
if k in s_name:
s_name = s_name.replace(k, "")
return s_name
def english_brand_recognition(self, standard_brand_name, s_name):
c_set = {'a':'', 'b':'', 'c':'', 'd':'', 'e':'', 'f':'', 'g':'', 'h':'', \
'i':'', 'j':'', 'k':'', 'l':'', 'm':'', 'n':'', 'o':'', 'p':'', \
'q':'', 'r':'', 's':'', 't':'', 'u':'', 'v':'', \
'w':'', 'x':'', 'y':'', 'z':''}
tmp_brand = None
if standard_brand_name in s_name:
lst1 = s_name.split(standard_brand_name)
for tmp in range(1, len(lst1)):
pre_str = lst1[tmp - 1]
next_str = lst1[tmp]
if pre_str == "" or next_str == "":
tmp_brand = standard_brand_name
break
else:
a, b = pre_str[-1], next_str[0]
if a not in c_set and b not in c_set:
tmp_brand = standard_brand_name
break
else:
tmp_brand = None
return tmp_brand
def brand_inclusion_relation_dealing(self, b_name_lst):
del_brand_dict = {}
for a in b_name_lst:
a1, _ = a.strip().lower().split("|")
for b in b_name_lst:
b1, _ = b.strip().lower().split("|")
if a1 == b1: continue
if a1 in b1: del_brand_dict[a] = ''
elif b1 in a1: del_brand_dict[b] = ''
else: continue
r_b_name_lst = []
for c in b_name_lst:
if c not in del_brand_dict:
r_b_name_lst.append(c)
return r_b_name_lst
def getting_high_gmv_brand(self, same_cat1_bid_lst):
tmp_gmv_lst = []
for y in same_cat1_bid_lst:
tmp_gmv_lst.append((self.brand_gmv_dict[y], self.name_ori_brand_dict[y], y))
tmp_gmv_lst = sorted(tmp_gmv_lst, key=lambda n: n[0], reverse=True)
return tmp_gmv_lst[0][1], tmp_gmv_lst[0][2]
def same_cat1_strategy(self, cat1_id, clean_brand_lst, clean_brand_id_lst):
pre_brand_id, pre_brand, match_type = None, None, None
same_id_lst = []
for a in clean_brand_lst:
same_id_lst += self.brand_idx_dict[a]
tmp_name_lst = []
for b in clean_brand_id_lst:
tmp_name_lst.append(self.name_ori_brand_dict[b])
same_cat1_bid_lst = []
for y in same_id_lst:
if cat1_id in self.brand_cat1_dict[y]:
pre_brand = self.name_ori_brand_dict[y]
pre_brand_id = y
same_cat1_bid_lst.append(y)
else:
pass
if len(same_cat1_bid_lst) == 1 and pre_brand != None:
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择相同一级类目的品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
# 相同一级类目，相同品牌名称，选择高GMV品牌
elif len(same_cat1_bid_lst) > 1:
pre_brand, pre_brand_id = self.getting_high_gmv_brand(same_cat1_bid_lst)
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择相同一级类目的高GMV品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
else:
pass
return pre_brand_id, pre_brand, match_type
def same_length_strategy(self, clean_brand_lst, clean_brand_id_lst):
same_id_lst = []
for a in clean_brand_lst:
same_id_lst += self.brand_idx_dict[a]
tmp_name_lst = []
for b in clean_brand_id_lst:
tmp_name_lst.append(self.name_ori_brand_dict[b])
if len(same_id_lst) == 1:
y = same_id_lst[0]
pre_brand = self.name_ori_brand_dict[y]
pre_brand_id = y
match_type = "1:匹配到多个不同品牌(%s)，2:选择最长品牌: %s" % ("|".join(tmp_name_lst), pre_brand)
else:
pre_brand,pre_brand_id = self.getting_high_gmv_brand(same_id_lst)
match_type = "1:匹配到多个相同最长度品牌名(%s)，2:选择相同一级类目的高GMV品牌: %s" % \
("|".join(tmp_name_lst), pre_brand)
return pre_brand_id, pre_brand, match_type
def same_cat1_gmv_strategy(self, cat1_id, clean_brand_lst, clean_brand_id_lst):
pre_brand_id, pre_brand, match_type = None, None, None
same_id_lst = []
for a in clean_brand_lst:
same_id_lst += self.brand_idx_dict[a]
tmp_name_lst = []
for b in clean_brand_id_lst:
tmp_name_lst.append(self.name_ori_brand_dict[b])
num = 0
for y in same_id_lst:
if cat1_id in self.brand_cat1_dict[y]:
pre_brand = self.name_ori_brand_dict[y]
pre_brand_id = y
num += 1
else:
pass
if num == 1 and pre_brand != None:
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择相同一级类目的品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
else:
pre_brand, pre_brand_id = self.getting_high_gmv_brand(same_id_lst)
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择高GMV品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
return pre_brand_id, pre_brand, match_type
def numerous_cat1_rule_func(self, b_id_lst):
d1 = {}
for z in b_id_lst:
if z not in self.brand_cat1_dict: continue
for y in self.brand_cat1_dict[z]:
if y in d1:
x = d1[y]
d1[y] = x + 1
else:
d1[y] = 1
lst1 = [(k, v) for k,v in d1.items()]
lst1 = sorted(lst1, key=lambda x: x[1], reverse=True)
c_id_1, c_num_1 = lst1[0]
if c_num_1 >= 3:
return False
'''
if len(lst1) >= 2:
c_id_2, c_num_2 = lst1[1]
if c_num_2 >=2 and c_num_1 >= 2:
return False
'''
return True
# def rule_opt(self, pre_brand_name_lst, cur_p_name, cur_cat1_id, cur_cat1_name,pre_clean_brand_id_lst):
# r_set = set()
# r_id_lst = []
# tmp_name_list = []
# r_name_list = []
#
# stp0_clean_lst = self.brand_rule_obj.phone_brand_not_appear_same_fun(pre_clean_brand_id_lst, cur_cat1_id)
#
# for tmp in list(stp0_clean_lst):
# if tmp not in self.idx_ori_brand_dict: continue
# tmp_name_list = tmp_name_list + self.brand_rule_obj.bid_2_bname(tmp)
# for b_name in tmp_name_list:
# b_name = b_name.lower()
# is_eng = tool.is_all_eng(b_name)
# flag = "0" if is_eng else "1"
# r_name_list.append("%s|%s" % (b_name, flag))
#
# stp1_clean_lst = self.brand_inclusion_relation_dealing(r_name_list)
# stp2_clean_lst = self.brand_rule_obj.co_appear_del_brand_func(stp1_clean_lst)
# stp3_clean_lst = self.brand_rule_obj.apppint_co_appear_del_brand_func(stp2_clean_lst)
# # stp2_clean_lst = self.brand_rule_obj.phone_brand_not_appear_same_fun(stp1_clean_lst, cur_cat1_id)
# stp4_clean_lst = self.brand_rule_obj.brand_not_appear_same_fun(stp3_clean_lst)
# for b_name in stp4_clean_lst:
# tmp_b_name, _ = b_name.split('|')
# # 修正过程。如“苹果手机支架”不为“苹果”手机，此时将“苹果”从候选列表中移除
# f1 = self.brand_rule_obj.brand_word_rule_func(tmp_b_name, cur_p_name)
# if f1 == False: continue
# f2 = self.brand_rule_obj.brand_cat1_rule_func(tmp_b_name, cur_cat1_name)
# if f2 == False: continue
# f3 = self.brand_rule_obj.brand_cat1_fixed_rule_func(tmp_b_name, cur_cat1_id)
# if f3 == False: continue
# r_set.add(b_name)
# r_id_lst += self.brand_idx_dict[b_name]
# r_b_name_lst, r_b_id_lst = list(r_set), list(set(r_id_lst))
# if len(r_b_id_lst) == 0:
# return [], []
# if not self.numerous_cat1_rule_func(r_b_id_lst):
# return [], []
# else:
# return r_b_name_lst, r_b_id_lst
def rule_opt(self, pre_brand_name_lst, cur_p_name, cur_cat1_id, cur_cat1_name):
r_set = set()
r_id_lst = []
stp0_clean_lst = self.brand_inclusion_relation_dealing(pre_brand_name_lst)
stp1_clean_lst = self.brand_rule_obj.co_appear_del_brand_func(stp0_clean_lst)
stp2_clean_lst = self.brand_rule_obj.apppint_co_appear_del_brand_func(stp1_clean_lst)
stp3_clean_lst = self.brand_rule_obj.phone_brand_not_appear_same_fun(stp2_clean_lst, cur_cat1_id)
stp4_clean_lst = self.brand_rule_obj.brand_not_appear_same_fun(stp3_clean_lst)
for b_name in stp4_clean_lst:
tmp_b_name, _ = b_name.split('|')
# 修正过程。如“苹果手机支架”不为“苹果”手机，此时将“苹果”从候选列表中移除
f1 = self.brand_rule_obj.brand_word_rule_func(tmp_b_name, cur_p_name)
if f1 == False: continue
f2 = self.brand_rule_obj.brand_cat1_rule_func(tmp_b_name, cur_cat1_name)
if f2 == False: continue
f3 = self.brand_rule_obj.brand_cat1_fixed_rule_func(tmp_b_name, cur_cat1_id)
if f3 == False: continue
r_set.add(b_name)
r_id_lst += self.brand_idx_dict[b_name]
r_b_name_lst, r_b_id_lst = list(r_set), list(set(r_id_lst))
if len(r_b_id_lst) == 0:
return [], []
if not self.numerous_cat1_rule_func(r_b_id_lst):
return [], []
else:
return r_b_name_lst, r_b_id_lst
def getting_cat1_info(self, brand_id):
brand_cat1_lst = self.brand_cat1_dict[brand_id]
brand_cat1_id = ",".join(brand_cat1_lst)
brand_cat1_name_lst = []
for yy in brand_cat1_lst:
brand_cat1_name_lst.append(self.cat1_dict[yy])
brand_cat1_name = ",".join(brand_cat1_name_lst)
return brand_cat1_id, brand_cat1_name
def brand_recognition(self, line_str):
'''
brand_recognition is updated by gcw in 2020.09.12.
'''
try:
line = line_str.strip()
if line == "": return None, None, None, None, None, None, None
lst_z = line.split("\001")
if len(lst_z) != 5: return None, None, None, None, None, None, None
lst1 = [tmp.strip() for tmp in lst_z]
product_id, ori_product_name, ori_brand_word, cat1_id, cat1_name = lst1
if product_id in self.brand_rule_obj.appoint_product_clean_dict:
brand_id = self.brand_rule_obj.appoint_product_clean_dict[product_id]
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(brand_id)
return brand_id, self.name_ori_brand_dict[brand_id], "匹配到唯一标准品牌", \
brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
# s_name小写字母
s_name = tool.s_name_dealing("%s %s" % (ori_product_name, ori_brand_word))
s_name = self.product_name_del_word(s_name)
clean_brand_lst = []
clean_brand_id_lst = []
# 1、清洗后的品牌
# 2、清洗后的品牌对应的多个品牌编号
for ori_b_str, same_brand_id_lst in self.brand_idx_dict.items():
# 品牌小写字母
b_str = ori_b_str.lower()
lst_z = b_str.split('|')
if len(lst_z) != 2: continue
b, is_eng = lst_z
if is_eng == "0" and tool.is_own_eng(s_name) and b in s_name:
tmp_b = self.english_brand_recognition(b, s_name)
if tmp_b != None:
clean_brand_lst.append(ori_b_str)
clean_brand_id_lst += same_brand_id_lst
elif b in s_name:
clean_brand_lst.append(ori_b_str)
clean_brand_id_lst += same_brand_id_lst
else:
continue
clean_brand_lst = list(set(clean_brand_lst))
clean_brand_id_lst = list(set(clean_brand_id_lst))
if len(clean_brand_lst) == 0: return None, None, "没有匹配到标准品牌", None, None, None, None
# 人为规则的过滤
clean_brand_lst, clean_brand_id_lst = self.rule_opt(clean_brand_lst, ori_product_name, cat1_id, cat1_name)
if len(clean_brand_lst) == 0:
return None, None, "没有匹配到标准品牌", None, None, None, None
if len(clean_brand_id_lst) == 1:
b_id = clean_brand_id_lst[0]
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(b_id)
return b_id, \
self.name_ori_brand_dict[clean_brand_id_lst[0]], \
"匹配到唯一标准品牌", brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
pre_brand_id, pre_brand, match_type = None, None, None
if len(clean_brand_lst) == 1:
pre_brand_id, pre_brand, match_type = \
self.same_cat1_gmv_strategy(cat1_id, clean_brand_lst, clean_brand_id_lst)
else:
len_brand_dict = {}
for tmp in clean_brand_lst:
l = len(tmp)
if l in len_brand_dict:
x = len_brand_dict[l]
len_brand_dict[l] = x + [tmp]
else:
len_brand_dict[l] = [tmp]
r_lst = [(k, v) for k, v in len_brand_dict.items()]
r_lst = sorted(r_lst, key=lambda m: m[0], reverse=True)
for u in r_lst:
tmp_l, tmp_clean_brand_lst = u
pre_brand_id, pre_brand, match_type = self.same_cat1_strategy(cat1_id, tmp_clean_brand_lst,
clean_brand_id_lst)
if pre_brand == None or pre_brand_id == None:
pre_brand_id, pre_brand, match_type = self.same_length_strategy(tmp_clean_brand_lst, clean_brand_id_lst)
if pre_brand != None and match_type != None:
break
if pre_brand_id != None:
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(pre_brand_id)
else:
brand_cat1_id = None
brand_cat1_name = None
return pre_brand_id, pre_brand, match_type, \
brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
except Exception as e:
raise e
def brand_same_cat1_recognition(self, line_str):
'''
brand_recognition is updated by gcw in 2020.09.12.
'''
try:
line = line_str.strip()
if line == "": return None, None, None, None, None, None, None
lst_z = line.split("\001")
if len(lst_z) != 5: return None, None, None, None, None, None, None
lst1 = [tmp.strip() for tmp in lst_z]
product_id, ori_product_name, ori_brand_word, cat1_id, cat1_name = lst1
# s_name小写字母
s_name = tool.s_name_dealing("%s %s" % (ori_product_name, ori_brand_word))
s_name = self.product_name_del_word(s_name)
clean_brand_lst = []
clean_brand_id_lst = []
# 1、清洗后的品牌
# 2、清洗后的品牌对应的多个品牌编号
for ori_b_str, same_brand_id_lst in self.brand_idx_dict.items():
# 品牌小写字母
b_str = ori_b_str.lower()
lst_z = b_str.split('|')
if len(lst_z) != 2: continue
b, is_eng = lst_z
if is_eng == "0" and tool.is_own_eng(s_name) and b in s_name:
tmp_b = self.english_brand_recognition(b, s_name)
if tmp_b != None:
clean_brand_lst.append(ori_b_str)
clean_brand_id_lst += same_brand_id_lst
elif b in s_name:
clean_brand_lst.append(ori_b_str)
clean_brand_id_lst += same_brand_id_lst
else:
continue
clean_brand_lst = list(set(clean_brand_lst))
clean_brand_id_lst = list(set(clean_brand_id_lst))
if len(clean_brand_lst) == 0: return None, None, "没有匹配到标准品牌", None, None, None, None
# 人为规则的过滤
clean_brand_lst, clean_brand_id_lst = self.rule_opt(clean_brand_lst, ori_product_name, cat1_id, cat1_name)
if len(clean_brand_lst) == 0:
return None, None, "没有匹配到标准品牌", None, None, None, None
if len(clean_brand_id_lst) == 1:
b_id = clean_brand_id_lst[0]
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(b_id)
return b_id, \
self.name_ori_brand_dict[clean_brand_id_lst[0]], \
"匹配到唯一标准品牌", brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
pre_brand_id, pre_brand, match_type = None, None, None
if len(clean_brand_lst) == 1:
pre_brand_id, pre_brand, match_type = \
self.same_cat1_gmv_strategy(cat1_id, clean_brand_lst, clean_brand_id_lst)
else:
len_brand_dict = {}
for tmp in clean_brand_lst:
l = len(tmp)
if l in len_brand_dict:
x = len_brand_dict[l]
len_brand_dict[l] = x + [tmp]
else:
len_brand_dict[l] = [tmp]
r_lst = [(k, v) for k, v in len_brand_dict.items()]
r_lst = sorted(r_lst, key=lambda m: m[0], reverse=True)
for u in r_lst:
tmp_l, tmp_clean_brand_lst = u
pre_brand_id, pre_brand, match_type = self.same_cat1_strategy(cat1_id, tmp_clean_brand_lst,
clean_brand_id_lst)
if pre_brand != None and match_type != None:
break
# 兜底处理
if pre_brand == None and match_type == None:
pre_brand_id, pre_brand, match_type = \
self.same_cat1_gmv_strategy(cat1_id, clean_brand_lst, clean_brand_id_lst)
if pre_brand_id != None:
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(pre_brand_id)
else:
brand_cat1_id = None
brand_cat1_name = None
return pre_brand_id, pre_brand, match_type, \
brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
except Exception as e:
raise e
class BrandRegTool(object):
def __init__(self,
standard_brand_file,
del_brand_file=None,
exchange_brand_file=None,
rule_brand_file=None):
if not os.path.exists(standard_brand_file):
raise Exception("%s does not exists!" % standard_brand_file)
try:
self.brand_loading_obj = BrandInfoLoading(standard_brand_file,
del_brand_file,
exchange_brand_file)
#直接读取进行过品牌扩展的文件
self.brand_idx_dict, self.idx_ori_brand_dict, self.name_ori_brand_dict, self.brand_cat1_dict, \
self.cat1_brand_dict, self.cat1_clean_brand_dict, self.brand_gmv_dict, \
self.cat1_dict = self.brand_loading_obj.brand_info_loading()
self.brand_rule_obj = BrandRefRuleOpt(rule_brand_file,
self.idx_ori_brand_dict)
except Exception as e:
raise e
def product_name_del_word(self, ori_s_name):
del_word_dict = {"网红":'', '美的布谷':'', '珂丝兰珂润万物丝兰完美':'', '淘米家用':'', \
'家用':'', 'OLOMLB':'', 'TESIRIS':'', '文和友演出门票':'', '淘米':'', \
'小米推荐':'', '小米众筹':'', '小米家人':'', '内蒙':'', '牛肉干':'', \
'小米诺':'', '小米琦':'', '小米粒':'', '小米奇':'','小米生态链':'', \
'小米有品同款':'', '小米同款':'', '正善牛肉哥':'', '东北大板栗':'', \
'黄小米':'', '大红印':'', '森马电商':'', '老板说了':'', '老板精选': ''}
s_name = ori_s_name
for k, v in del_word_dict.items():
if k in s_name:
s_name = s_name.replace(k, "")
return s_name
def english_brand_recognition(self, standard_brand_name, s_name):
c_set = {'a':'', 'b':'', 'c':'', 'd':'', 'e':'', 'f':'', 'g':'', 'h':'', \
'i':'', 'j':'', 'k':'', 'l':'', 'm':'', 'n':'', 'o':'', 'p':'', \
'q':'', 'r':'', 's':'', 't':'', 'u':'', 'v':'', \
'w':'', 'x':'', 'y':'', 'z':''}
tmp_brand = None
if standard_brand_name in s_name:
lst1 = s_name.split(standard_brand_name)
for tmp in range(1, len(lst1)):
pre_str = lst1[tmp - 1]
next_str = lst1[tmp]
if pre_str == "" or next_str == "":
tmp_brand = standard_brand_name
break
else:
a, b = pre_str[-1], next_str[0]
if a not in c_set and b not in c_set:
tmp_brand = standard_brand_name
break
else:
tmp_brand = None
return tmp_brand
def brand_inclusion_relation_dealing(self, b_name_lst):
del_brand_dict = {}
for a in b_name_lst:
a1, _ = a.strip().lower().split("|")
for b in b_name_lst:
b1, _ = b.strip().lower().split("|")
if a1 == b1: continue
if a1 in b1: del_brand_dict[a] = ''
elif b1 in a1: del_brand_dict[b] = ''
else: continue
r_b_name_lst = []
for c in b_name_lst:
if c not in del_brand_dict:
r_b_name_lst.append(c)
return r_b_name_lst
def getting_high_gmv_brand(self, same_cat1_bid_lst):
tmp_gmv_lst = []
for y in same_cat1_bid_lst:
tmp_gmv_lst.append(
(self.brand_gmv_dict[y], self.name_ori_brand_dict[y], y))
tmp_gmv_lst = sorted(tmp_gmv_lst, key=lambda n: n[0], reverse=True)
return tmp_gmv_lst[0][1], tmp_gmv_lst[0][2]
def same_cat1_strategy(self, cat1_id, clean_brand_lst, clean_brand_id_lst):
pre_brand_id, pre_brand, match_type = None, None, None
same_id_lst = []
for a in clean_brand_lst:
same_id_lst += self.brand_idx_dict[a]
tmp_name_lst = []
for b in clean_brand_id_lst:
tmp_name_lst.append(self.name_ori_brand_dict[b])
same_cat1_bid_lst = []
for y in same_id_lst:
if cat1_id in self.brand_cat1_dict[y]:
pre_brand = self.name_ori_brand_dict[y]
pre_brand_id = y
same_cat1_bid_lst.append(y)
else:
pass
if len(same_cat1_bid_lst) == 1 and pre_brand != None:
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择相同一级类目的品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
# 相同一级类目，相同品牌名称，选择高GMV品牌
elif len(same_cat1_bid_lst) > 1:
pre_brand, pre_brand_id = self.getting_high_gmv_brand(
same_cat1_bid_lst)
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择相同一级类目的高GMV品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
else:
pass
return pre_brand_id, pre_brand, match_type
def same_length_strategy(self, clean_brand_lst, clean_brand_id_lst):
same_id_lst = []
for a in clean_brand_lst:
same_id_lst += self.brand_idx_dict[a]
tmp_name_lst = []
for b in clean_brand_id_lst:
tmp_name_lst.append(self.name_ori_brand_dict[b])
if len(same_id_lst) == 1:
y = same_id_lst[0]
pre_brand = self.name_ori_brand_dict[y]
pre_brand_id = y
match_type = "1:匹配到多个不同品牌(%s)，2:选择最长品牌: %s" % (
"|".join(tmp_name_lst), pre_brand)
else:
pre_brand, pre_brand_id = self.getting_high_gmv_brand(same_id_lst)
match_type = "1:匹配到多个相同最长度品牌名(%s)，2:选择相同一级类目的高GMV品牌: %s" % \
("|".join(tmp_name_lst), pre_brand)
return pre_brand_id, pre_brand, match_type
def same_cat1_gmv_strategy(self, cat1_id, clean_brand_lst,
clean_brand_id_lst):
pre_brand_id, pre_brand, match_type = None, None, None
same_id_lst = []
for a in clean_brand_lst:
same_id_lst += self.brand_idx_dict[a]
tmp_name_lst = []
for b in clean_brand_id_lst:
tmp_name_lst.append(self.name_ori_brand_dict[b])
num = 0
for y in same_id_lst:
if cat1_id in self.brand_cat1_dict[y]:
pre_brand = self.name_ori_brand_dict[y]
pre_brand_id = y
num += 1
else:
pass
if num == 1 and pre_brand != None:
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择相同一级类目的品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
else:
pre_brand, pre_brand_id = self.getting_high_gmv_brand(same_id_lst)
match_type = "1:匹配到相同品牌名(%s)多个不同品牌(%s)，2:选择高GMV品牌" % \
(clean_brand_lst[0], "|".join(tmp_name_lst))
return pre_brand_id, pre_brand, match_type
def same_cat1_rule_func(self, b_id_lst, cur_cat1_id):
'''
仅仅保留相同一级类目的品牌
:param b_id_lst:
:param cur_cat1_id:
:return:
'''
r_lst = []
for z in b_id_lst:
if z not in self.brand_cat1_dict: continue
ok = self.brand_cat1_dict[z]
if cur_cat1_id in self.brand_cat1_dict[z]:
r_lst.append(z)
else:
continue
return r_lst
def rule_opt(self, pre_brand_name_lst, pre_idx_2_brand_dict,\
cur_p_name, cur_cat1_id, cur_cat1_name):
r_set = set()
r_id_lst = []
stp2_clean_lst = self.brand_rule_obj.co_appear_del_brand_func(
pre_brand_name_lst)
stp3_clean_lst = self.brand_rule_obj.apppint_co_appear_del_brand_func(
stp2_clean_lst)
stp4_clean_lst = self.brand_rule_obj.brand_not_appear_same_fun(
stp3_clean_lst)
# 有问题
for b_name in stp4_clean_lst:
tmp_b_name, _ = b_name.split('|')
# 修正过程。如“苹果手机支架”不为“苹果”手机，此时将“苹果”从候选列表中移除
f1 = self.brand_rule_obj.brand_word_rule_func(
tmp_b_name, cur_p_name, cur_cat1_id)
if f1 == False: continue
f2 = self.brand_rule_obj.brand_cat1_rule_func(
tmp_b_name, cur_cat1_name)
if f2 == False: continue
f3 = self.brand_rule_obj.brand_cat1_fixed_rule_func(
tmp_b_name, cur_cat1_id)
if f3 == False: continue
r_set.add(b_name)
r_id_lst += self.brand_idx_dict[b_name]
r_b_name_lst, r_b_id_lst = list(r_set), list(set(r_id_lst))
if len(r_b_id_lst) == 0: return [], []
# 人工强制规定一个品牌规则
r_b_name_lst, r_b_id_lst = self.manual_compellent_assign_brand_rule(
r_b_name_lst, r_b_id_lst)
r_b_id_lst = self.same_cat1_rule_func(r_b_id_lst, cur_cat1_id)
r_b_id_lst = self.brand_rule_obj.phone_brand_not_appear_same_fun(
r_b_id_lst, cur_cat1_id)
r_b_name_lst = []
for xx in r_b_id_lst:
if xx not in pre_idx_2_brand_dict: continue
r_b_name_lst += pre_idx_2_brand_dict[xx]
if len(r_b_id_lst) == 0:
return [], []
else:
r_b_name_lst = []
for xx in r_b_id_lst:
if xx not in pre_idx_2_brand_dict: continue
r_b_name_lst += pre_idx_2_brand_dict[xx]
return r_b_name_lst, r_b_id_lst
def getting_cat1_info(self, brand_id):
brand_cat1_lst = self.brand_cat1_dict[brand_id]
brand_cat1_id = ",".join(brand_cat1_lst)
brand_cat1_name_lst = []
for yy in brand_cat1_lst:
brand_cat1_name_lst.append(self.cat1_dict[yy])
brand_cat1_name = ",".join(brand_cat1_name_lst)
return brand_cat1_id, brand_cat1_name
def manual_compellent_assign_brand_rule(self, r_b_name_lst, r_b_id_lst):
'''
解决了 【资生堂】【安热沙】【安耐晒】三个同时出现是，品牌为：【安热沙】
例如：3428308113219031576
解决了 【资生堂】【安热沙】两个同时出现时，品牌为：【安热沙】
例如：3429766112856943278
解决了 【小米有品】【麦瑞克】两个同时出现时，品牌为：【麦瑞克】
例如：3429256598957224027,3432639012387570550
'''
if '10266841' in r_b_id_lst and '12661143' in r_b_id_lst and '10756319' in r_b_id_lst and len(
r_b_id_lst) == 3:
return ['安热沙|0'], ['10266841']
if '10266841' in r_b_id_lst and '10756319' in r_b_id_lst and len(
r_b_id_lst) == 2:
return ['安热沙|0'], ['10266841']
elif '10698337' in r_b_id_lst and '10032446' in r_b_id_lst and len(
r_b_id_lst) == 2:
return ['麦瑞克|0'], ['10032446']
else:
return r_b_name_lst, r_b_id_lst
def brand_recognition(self, line_str):
'''
brand_recognition is updated by gcw in 2020.09.12.
'''
try:
line = line_str.strip()
if line == "": return None, None, None, None, None, None, None
lst_z = line.split("\001")
if len(lst_z) != 5: return None, None, None, None, None, None, None
lst1 = [tmp.strip() for tmp in lst_z]
product_id, ori_product_name, _, cat1_id, cat1_name = lst1
# s_name小写字母
s_name = tool.s_name_dealing(ori_product_name)
s_name = self.product_name_del_word(s_name)
clean_brand_2_idx = {}
clean_brand_lst = []
# 英语手机品牌清洗，比如：vivo，oppo等的清洗
englisg_phone_brand_lst = self.brand_rule_obj.english_phone_rule_func(
s_name, cat1_id)
for epb in englisg_phone_brand_lst:
epb_id, epb_name = epb
epb_name = epb_name + '|0'
if epb_name not in self.brand_idx_dict: continue
clean_brand_2_idx[epb_name] = epb_id
clean_brand_lst.append(epb_name)
# 1、清洗后的品牌
# 2、清洗后的品牌对应的多个品牌编号
for ori_b_str, same_brand_id_lst in self.brand_idx_dict.items():
# 品牌小写字母
b_str = ori_b_str.lower()
lst_z = b_str.split('|')
if len(lst_z) != 2: continue
b, is_eng = lst_z
if is_eng == "0" and tool.is_own_eng(s_name) and b in s_name:
tmp_b = self.english_brand_recognition(b, s_name)
if tmp_b != None:
for tmp_bid in same_brand_id_lst:
clean_brand_2_idx[ori_b_str] = tmp_bid
clean_brand_lst.append(ori_b_str)
elif b in s_name:
for tmp_bid in same_brand_id_lst:
clean_brand_2_idx[ori_b_str] = tmp_bid
clean_brand_lst.append(ori_b_str)
else:
continue
# 去除已识别品牌中的包含关系
clean_brand_lst = self.brand_inclusion_relation_dealing(
clean_brand_lst)
clean_idx_2_brand = {}
for xxx in clean_brand_lst:
x_id = clean_brand_2_idx[xxx]
if x_id in clean_brand_2_idx:
zzz = clean_idx_2_brand[x_id]
clean_idx_2_brand[x_id] = zzz + [xxx]
else:
clean_idx_2_brand[x_id] = [xxx]
if len(clean_brand_lst) == 0:
return None, None, "没有匹配到标准品牌", None, None, None, None
# 人为规则的过滤
clean_brand_lst, clean_brand_id_lst = self.rule_opt(clean_brand_lst, clean_idx_2_brand, \
ori_product_name, cat1_id, cat1_name)
if len(clean_brand_lst) == 0:
return None, None, "没有匹配到标准品牌", None, None, None, None
if len(clean_brand_id_lst) == 1:
b_id = clean_brand_id_lst[0]
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(b_id)
return b_id, \
self.name_ori_brand_dict[clean_brand_id_lst[0]], \
"匹配到唯一标准品牌", brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
pre_brand_id, pre_brand, match_type = None, None, None
if len(clean_brand_lst) == 1:
pre_brand_id, pre_brand, match_type = \
self.same_cat1_gmv_strategy(cat1_id, clean_brand_lst, clean_brand_id_lst)
else:
len_brand_dict = {}
for tmp in clean_brand_lst:
l = len(tmp)
if l in len_brand_dict:
x = len_brand_dict[l]
len_brand_dict[l] = x + [tmp]
else:
len_brand_dict[l] = [tmp]
r_lst = [(k, v) for k, v in len_brand_dict.items()]
r_lst = sorted(r_lst, key=lambda m: m[0], reverse=True)
for u in r_lst:
tmp_l, tmp_clean_brand_lst = u
pre_brand_id, pre_brand, match_type = self.same_cat1_strategy(
cat1_id, tmp_clean_brand_lst, clean_brand_id_lst)
if pre_brand == None or pre_brand_id == None:
pre_brand_id, pre_brand, match_type = self.same_length_strategy(
tmp_clean_brand_lst, clean_brand_id_lst)
if pre_brand != None and match_type != None:
break
if pre_brand_id != None:
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(
pre_brand_id)
else:
brand_cat1_id = None
brand_cat1_name = None
return pre_brand_id, pre_brand, match_type, \
brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
except Exception as e:
raise e
def brand_reg_opt(self, reg_brand_dict, line_str):
try:
line = line_str.strip()
if line == "": return None, None, None, None, None, None, None
lst_z = line.split("\001")
if len(lst_z) != 5: return None, None, None, None, None, None, None
lst1 = [tmp.strip() for tmp in lst_z]
product_id, ori_product_name, ori_brand_word, cat1_id, cat1_name = lst1
#
# s_name小写字母
s_name = tool.s_name_dealing("%s %s" %
(ori_product_name, ori_brand_word))
s_name = self.product_name_del_word(s_name)
clean_idx_2_brand = {}
# 英语手机品牌清洗，比如：vivo，oppo等的清洗
englisg_phone_brand_lst = self.brand_rule_obj.english_phone_rule_func(
s_name, cat1_id)
for epb in englisg_phone_brand_lst:
epb_id, epb_name = epb
epb_name = epb_name + '|0'
if epb_name not in self.brand_idx_dict: continue
clean_idx_2_brand[epb_id] = epb_name
# 1、清洗后的品牌
# 2、清洗后的品牌对应的多个品牌编号
for ori_b_str, same_brand_id_lst in reg_brand_dict.items():
# 品牌小写字母
b_str = ori_b_str.lower()
lst_z = b_str.split('|')
if len(lst_z) != 2: continue
b, is_eng = lst_z
if is_eng == "0" and tool.is_own_eng(s_name) and b in s_name:
tmp_b = self.english_brand_recognition(b, s_name)
if tmp_b != None:
for tmp_bid in same_brand_id_lst:
clean_idx_2_brand[tmp_bid] = ori_b_str
elif b in s_name:
for tmp_bid in same_brand_id_lst:
clean_idx_2_brand[tmp_bid] = ori_b_str
else:
continue
clean_brand_lst = list(set(list(clean_idx_2_brand.values())))
clean_brand_id_lst = list(set(list(clean_idx_2_brand.keys())))
if len(clean_brand_lst) == 0:
return None, None, "没有匹配到标准品牌", None, None, None, None
# 人为规则的过滤
clean_brand_lst, clean_brand_id_lst = self.rule_opt(
clean_brand_id_lst, clean_idx_2_brand, ori_product_name,
cat1_id, cat1_name)
if len(clean_brand_lst) == 0:
return None, None, "没有匹配到标准品牌", None, None, None, None
if len(clean_brand_id_lst) == 1:
b_id = clean_brand_id_lst[0]
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(b_id)
return b_id, \
self.name_ori_brand_dict[clean_brand_id_lst[0]], \
"匹配到唯一标准品牌", brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
pre_brand_id, pre_brand, match_type = None, None, None
if len(clean_brand_lst) == 1:
pre_brand_id, pre_brand, match_type = \
self.same_cat1_gmv_strategy(cat1_id, clean_brand_lst, clean_brand_id_lst)
else:
len_brand_dict = {}
for tmp in clean_brand_lst:
l = len(tmp)
if l in len_brand_dict:
x = len_brand_dict[l]
len_brand_dict[l] = x + [tmp]
else:
len_brand_dict[l] = [tmp]
r_lst = [(k, v) for k, v in len_brand_dict.items()]
r_lst = sorted(r_lst, key=lambda m: m[0], reverse=True)
for u in r_lst:
tmp_l, tmp_clean_brand_lst = u
pre_brand_id, pre_brand, match_type = self.same_cat1_strategy(
cat1_id, tmp_clean_brand_lst, clean_brand_id_lst)
if pre_brand == None or pre_brand_id == None:
pre_brand_id, pre_brand, match_type = self.same_length_strategy(
tmp_clean_brand_lst, clean_brand_id_lst)
if pre_brand != None and match_type != None:
break
if pre_brand_id != None:
brand_cat1_id, brand_cat1_name = self.getting_cat1_info(
pre_brand_id)
else:
brand_cat1_id = None
brand_cat1_name = None
return pre_brand_id, pre_brand, match_type, \
brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
except Exception as e:
raise e
def brand_same_cat1_recognition(self, line_str):
'''
brand_recognition is updated by gcw in 2020.09.12.
'''
try:
line = line_str.strip()
if line == "": return None, None, None, None, None, None, None
lst_z = line.split("\001")
if len(lst_z) != 5: return None, None, None, None, None, None, None
lst1 = [tmp.strip() for tmp in lst_z]
product_id, ori_product_name, ori_brand_word, ori_cat1_id, ori_cat1_name = lst1
pre_brand_id, pre_brand, match_type, \
brand_cat1_id, brand_cat1_name, cat1_id, cat1_name = None, None, None, None, None, None, None
reg_brand_dict = {}
if ori_cat1_id in self.cat1_clean_brand_dict:
brand_lst = self.cat1_clean_brand_dict[ori_cat1_id]
for tmp_b in brand_lst:
if tmp_b not in self.brand_idx_dict: continue
reg_brand_dict[tmp_b] = self.brand_idx_dict[tmp_b]
#
pre_brand_id, pre_brand, match_type, brand_cat1_id, brand_cat1_name, \
cat1_id, cat1_name = self.brand_reg_opt(reg_brand_dict, line)
if pre_brand == None or pre_brand_id == None:
pre_brand_id, pre_brand, match_type, brand_cat1_id, brand_cat1_name, \
cat1_id, cat1_name = self.brand_reg_opt(self.brand_idx_dict, line)
return pre_brand_id, pre_brand, match_type, brand_cat1_id, brand_cat1_name, cat1_id, cat1_name
except Exception as e:
raise e