添加品牌列的黑名单

predict/base_handler.py

@@ -290,6 +290,6 @@ class BasePredictor(metaclass=ClassBasePredictorMeta):
         return round(count / len(data), 3) >= SIMPLE_ENCAP_LEVEL
 
 if __name__ == "__main__":
-    print(BasePredictor.is_num(['数量', 1000]))
+    print(BasePredictor.is_catecol(['刁妍月 TK17091576628-10套   客供 单面 14.2*26.2   PCB订单号：588259  顺丰寄付 X-RAY检测  气泡袋包装', '种类', 'TRA', 'RES', 'RES', 'RES', 'RES', 'RES', 'RES', 'RES', 'RES', 'RES', 'RES', 'RES', 'OTH', 'MAG', 'IC', 'IC', 'IC', 'IC', 'IC', 'IC', 'IC', 'IC', 'IC', 'IC', 'DIO', 'DIO', 'DIO', 'DIO', 'CAP', 'CAP', 'CAP', 'CAP', 'CAP', 'CAP', 'CAP']))
 
     # print(re.match(r'(\d+|\d+(\.\d+))($|(K)|[个十百千万])$', '1k', re.M | re.I))
\ No newline at end of file

predict/dict_predict.py

@@ -200,16 +200,23 @@ class DicPredict(BasePredictor):
             for col_nullrate in sort_li_nullrate[1:]:
                 temp_pre_model_res.pop(col_nullrate[0])
 
-        # 若有多个品牌列,则进行空置率的比较, 空置率最低的选为目标列
+        # 模型对于品牌的识别不太精准, 若出现有多个品牌列, 先进行黑名单词汇的过滤, 若仍有多列最后再进行空置率的比较, 空置率最低的选为目标列
         prob_brand_cols = [i for i in temp_pre_model_res if temp_pre_model_res[i] == '品牌']
         if len(prob_brand_cols) >= 2:
-            li_nullrate = []
             for prob_brand_col in prob_brand_cols:
-                nanrate = BasePredictor.get_nan_rate(dic_data[prob_brand_col])
-                li_nullrate.append((prob_brand_col, nanrate))
-            sort_li_nullrate = sorted(li_nullrate, key=lambda x: x[1])
-            for col_nullrate in sort_li_nullrate[1:]:
-                temp_pre_model_res.pop(col_nullrate[0])
+                for i in dic_data[prob_brand_col]:
+                    if str(i).strip().lower() in BRAND_BLACK_LIST:
+                        temp_pre_model_res.pop(prob_brand_col)
+                        break
+            filter_brand_black_cols = [i for i in temp_pre_model_res if temp_pre_model_res[i] == '品牌']
+            if len(filter_brand_black_cols) >= 2:
+                li_nullrate = []
+                for prob_brand_col in filter_brand_black_cols:
+                    nanrate = BasePredictor.get_nan_rate(dic_data[prob_brand_col])
+                    li_nullrate.append((prob_brand_col, nanrate))
+                sort_li_nullrate = sorted(li_nullrate, key=lambda x: x[1])
+                for col_nullrate in sort_li_nullrate[1:]:
+                    temp_pre_model_res.pop(col_nullrate[0])
 
         # 如果没有数量列则把之前的序号列(其实是数字列)当成数字列
         if not prob_num_cols:

static_config.py

 #!/usr/bin/env python
 # -*- coding:utf-8 -*-
 # 标准名和代名词的映射
-li_category = ["类别", "分类", "名称", "类别名称", "类型", "产品分类"]
+li_category = ["类别", "分类", "名称", "类别名称", "类型", "产品分类", "种类"]
 li_param = ["参数", "规格", "描述", "值", "description"]
 li_gn = ["型号", "参考料号", "料号", "mpn", "厂商编码", "元器件", "规格型号"]
 li_num = ["数量", "用量(pcs)", "用量", "pcs", "quantity", "qty", "buy qty", "buy quantity", "单板数量", "采购数量", "单机用量", "单机数量", "单板用量"]
-li_brand = ["品牌", "品牌/厂商", "参考品牌", "参考供应商", "厂商", "参考供应商", "参考厂商", "manufacturer制造商", "manufacturer", "厂牌"]
+li_brand = ["品牌", "品牌/厂商", "参考品牌", "参考供应商", "厂商", "参考供应商", "参考厂商", "manufacturer制造商", "manufacturer", "厂牌", "公司"]
 li_encap = ["封装", "封装规格", "encapsulation", "footprint封装", "packagereference"]
 li_position = ["位号", "位置", "标号", "点位"]
 
@@ -27,7 +27,7 @@ li_position.extend(['*' + i for i in li_position])
 
 
 
-PROB_FIELDS = ["序号", "a面位置", "b面位置", "备注", "售价", "item", "top面", "bottom面", "designator", "remark"]
+PROB_FIELDS = ["序号", "a面位置", "b面位置", "备注", "售价", "item", "top面", "bottom面", "designator", "remark", "工艺"]
 AB_FIELDS = PROB_FIELDS + ['*' + i for i in PROB_FIELDS]
 # 可能的头部字段
 ALL_FIELDS = AB_FIELDS + li_category + li_param + li_gn + li_num + li_brand + li_encap + li_position
@@ -61,7 +61,10 @@ CATEGORY = ["半导体", "嵌入式", "光电子", "光源", "无源", "连接
             "铰接件", "盖塞与遮蔽", "支撑脚和轮", "型材与座", "减震器", "轴承", "把手", "直线导轨", "技术密封", "切换夹具和夹紧螺栓", "边缘保护套", "磁铁", "报警器",
             "分线盒", "变送器", "调制与控制", "气动", "螺线管", "启动器", "阀门和汇流板", "空气设备", "仪表", "气动配件", "测量", "软焊和焊接设备", "PCB原型制作与生产",
             "化学制剂", "工具", "防静电保护", "健康与安全", "收音机和CB用框架及半框架", "扬声器罩，前面板固定座", "排线", "电机控制模块", "微型电动机", "底板", "3D打印机",
-            "RTV服务", "配件", "三极管", '钽电容', 'MOS管', '芯片', '器件', '模组', '接线', '按键', 'IC', 'LDO']
+            "RTV服务", "配件", "三极管", '钽电容', 'MOS管', '芯片', '器件', '模组', '接线', '按键', 'IC', 'LDO', 'DIO', 'CAP', 'RES', 'TRA']
+
+# 模型识别黑名单词汇列表, 在可能目标列大于等于2的时候使用
+BRAND_BLACK_LIST = ['smt', 'ic', 'res']
 
 # 冲突时识别优先级,暂时没用
 DIC_PRIORITY = {