完善项目名标准化

api/constants.py

@@ -1,12 +1,24 @@
 # constants.py
-#匹配工程名时，需要过滤掉的词汇，线路， "变电站","换流站","变电",","调试部分"
-USELESS_PROJECT_WORDS = ["项目", "工程", "变电站", "线路", "变电","千伏" ,"换流站","公司","直流","部分","施工"]
-# USELESS_PROJECT_WORDS = ["项目", "工程", "千伏" ,"公司","直流"]
 
-#匹配公司名时，需要过滤掉的词汇
+#提取工程名热词中的地名时需要用的到过滤词列表
+# USELESS_PROJECT_WORDS = {"项目", "工程", "变电站", "线路", "变电", "千伏", "换流站", "公司", "直流", "部分", "施工",
+#                          "电厂", "牵引站","改造","间隔","维修","中心","主设备","应急","抢修","服务","扩建","完善","新建",
+#                            "配电","装置","调试","显示器","维护","设备","检修","电缆","光伏","保护","站","国网","安徽","阜阳",
+#                          "架空","电气安装包","安装包","主变"}
+#工程名标准化时需要过滤掉的词汇
+USELESS_PROJECT_WORDS = ["项目", "工程", "千伏", "公司", "直流"]
+
+#项目名标准化时需要过滤掉的词汇
+USELESS_PROGRAM_DEPARTMENT_WORDS = {"项目管理部","项目部", "项目", "管理"}
+
+#公司名标准化时需要过滤掉的词汇
 USELESS_COMPANY_WORDS = ["公司","有限","责任","工程","科技"]
 
-COMPANYNAME_SHA = "顺安电网建设有限公司"
+#提取公司名热词需要过滤掉的词汇
+# USELESS_COMPANY_WORDS = ["公司","有限","责任","工程","科技","安徽省","国网","四川省","安徽","集团","电力","建设","建筑","安装","股份"
+#                          "装饰","结构","能源","发展","装饰","电气","股份"]
+
+#槽位抽取时各槽位字段名
 #日期
 DATE = "date"
 #工程名称

api/main.py

@@ -18,8 +18,8 @@ from apscheduler.schedulers.background import BackgroundScheduler
 # MODEL_ERNIE_PATH = R"../ernie/output/checkpoint-25910"
 # MODEL_UIE_PATH = R"../uie/output/checkpoint-32750"
 
-MODEL_ERNIE_PATH = R"../ernie/output_temp/checkpoint-33510"
+MODEL_ERNIE_PATH = R"../ernie/output_temp/checkpoint-34340"
-MODEL_UIE_PATH = R"../uie/output_temp/checkpoint-33220"
+MODEL_UIE_PATH = R"../uie/output_temp/checkpoint-34050"
 
 # 类别名称列表
 labels = [
@@ -283,7 +283,7 @@ def extract_multi_chat(messages):
 
     logger.info(f"len(history_messages):{len(history_messages)}")
 
-    #最新问题的上一个问题里如何含有时间，则清空最老的历史对话
+    #最新问题的上一个问题里如果含有时间，则清空最老的历史对话
     last_two_messages = history_messages[-2:]
     has_time_prefix = any(
         msg.role == "user" and any(prefix in msg.content and prefix != msg.content for prefix in time_prefixes)

api/utils.py

@@ -12,7 +12,7 @@ import re
 
 from globalData import GlobalData
 from constants import USELESS_COMPANY_WORDS, USELESS_PROJECT_WORDS, CONSTRUCTION_UNIT, IMPLEMENTATION_ORG, \
-    SUBCONTRACTOR, PROJECT_NAME, PROJECT_DEPARTMENT, RISK_LEVEL, TEAM_NAME
+    SUBCONTRACTOR, PROJECT_NAME, PROJECT_DEPARTMENT, RISK_LEVEL, TEAM_NAME, USELESS_PROGRAM_DEPARTMENT_WORDS
 
 from logger_util import setup_logger
 
@@ -312,16 +312,23 @@ def standardize_projectDepartment(standard_company, input_project, company_proje
     """
     try:
         # **2. 先尝试直接匹配最相似的项目名**
-        input_project = replace_arabic_with_chinese(input_project)
+        temp_input_project = replace_arabic_with_chinese(input_project)
-        project_match = process.extractOne(input_project, company_project_department_map[standard_company],
+
+        temp_input_project = clean_useless_program_departement_name(temp_input_project)
+
+        program_list = company_project_department_map.get(standard_company, [])
+        cleaned_map = {clean_useless_program_departement_name(p): p for p in program_list}
+
+        project_match = process.extractOne(temp_input_project, list(cleaned_map.keys()),
                                            scorer=cast(Callable, WRatio))
-        logger.info(f"项目部名称最相似：{project_match[0]},{project_match[1]}")
+
-        # print(f"项目部名称最相似：{project_match[0]},{project_match[1]}", flush=True)
+        logger.info(f"{input_project} 名称最相似：{project_match[0]}, {project_match[1]}")
         if project_match and project_match[1] >= high_score:
-            return [project_match[0]]  # 直接返回匹配的项目名
+            matched_raw = cleaned_map[project_match[0]]
+            return [matched_raw]  # 返回原始名称
 
         # **3. 提取项目部的数字部分**
-        query_number = extract_number(input_project)
+        query_number = extract_number(temp_input_project)
 
         # **4. 过滤所有符合数字的项目部**
         matched_projects = []
@@ -356,8 +363,6 @@ def multiple_standardize_single_name(origin_input_name, origin_name_list, pinyin
                                     limit=len(origin_name_list))
     # 找到所有相似度 > 80 的匹配项
     original_high_confidence_matches = [(match[0], match[1]) for match in match_results if match[1] >= lower_score]
-    # print(f"standardize_pinyin_single_name 原始名匹配, high_confidence_matches:{original_high_confidence_matches[:3]}",
-    #       flush=True)
     logger.info(f"standardize_pinyin_single_name 原始名匹配, high_confidence_matches:{original_high_confidence_matches[:3]}")
 
     combined_low_confidence_matches = []
@@ -386,7 +391,6 @@ def generate_project_prompt_with_key(matched_projects, original_name="", slot_ke
     返回:
         str: 生成的提示信息。如果未找到匹配项，返回提示用户提供更准确信息的字符串。
     """
-    # print(f"generate_project_prompt_with_key slot_key:{slot_key},original_name:{original_name},matched_projects:{matched_projects} ")
     logger.info(f"generate_project_prompt_with_key slot_key:{slot_key},original_name:{original_name},matched_projects:{matched_projects} ")
     type = ""
     if slot_key == CONSTRUCTION_UNIT:
@@ -480,6 +484,8 @@ class StandardType(Enum):
 # 构建一个用于替换的正则表达式
 useless_project_words_pattern = re.compile("|".join(USELESS_PROJECT_WORDS))
 useless_company_words_pattern = re.compile("|".join(USELESS_COMPANY_WORDS))
+useless_program_department_words_pattern = re.compile("|".join(USELESS_PROGRAM_DEPARTMENT_WORDS))
+
 # 匹配所有数字、字母（含大小写）、特殊字符（包括空格、标点）
 project_symbols_pattern = re.compile(r"[A-Za-z0-9\s\W_ⅰ-ⅷⅠ-Ⅻⅸⅹⅺⅻ]+", re.UNICODE)
 #特殊字符
@@ -507,6 +513,14 @@ def clean_useless_team_leader_name(name: str) -> str:
     name = useless_team_leader_words_pattern.sub("", name)
     return name.strip()
 
+#去掉项目部里面的不重要词汇
+def clean_useless_program_departement_name(name: str) -> str:
+    # 去掉无意义词
+    name = useless_program_department_words_pattern.sub("", name)
+    # 去掉数字、字母、符号
+    name = project_symbols_pattern.sub("", name)
+    return name.strip()
+
 #槽位缺失检查
 def check_lost(int_res, slot):
     #labels: ["天气查询","通用对话","页面切换","日计划数量查询","周计划数量查询","日计划作业内容","周计划作业内容","施工人数","作业考勤人数","知识问答"]

generated_data/generated.py

@@ -57,7 +57,7 @@ BASE_DATA = {
                             "第三项目管理部（谯城变、亳州楼）","第十项目管理部（亳州变电）","第三项目管理部香鹭东段","第三项目部六安线路","第八项目管理部（宿州分部）",
                             "第八项目管理部宿州分部"],
     # 项目经理
-    "project_managers": ["陈少平项目经理", "范文立项目经理", "何东洋项目经理"],
+    "project_managers": ["陈少平项目经理", "范文立项目经理", "何东洋项目经理","周易开项目经理"],
 
     # 建管地区，"国网安徽省电力有限公司建设分公司", "国网安徽省电力有限公司马鞍山供电公司","马鞍山供电公司",
 
@@ -69,7 +69,7 @@ BASE_DATA = {
     "subcontractors": ["劦力建筑责任公司","安徽劦力建筑装饰有限责任公司", "安徽苏亚建设集团有限公司","大信电力建设有限公司","优越电力公司",
                        "安徽国腾电力工程有限公司","安徽京硚建设有限公司","中国能源建设集团安徽省电力设计院有限公司"],
     # 班组名称
-    "team_names": ["张朵班组", "刘梁玉班组", "魏玉龙班组"],
+    "team_names": ["张朵班组", "刘梁玉班组", "魏玉龙班组","周可富班组"],
     # 班组长
     "team_leaders": ["李元帅班组长", "刘雨豪班组长"],
     # 风险等级
@@ -744,21 +744,21 @@ TEMPLATE_CONFIG = {
         "templates": [
             ("打开{page}页面", ["page"]),
             ("打开{page}", ["page"]),
-            ("打开{page}模块", ["page"]),
+            ("切换{page}模块", ["page"]),
             ("进入{page}", ["page"]),
             ("进入{page}模块", ["page"]),
-            ("进入{page}页面", ["page"]),
+            ("切换到{page}页面", ["page"]),
             ("跳转到{page}", ["page"]),
             ("跳转到{page}模块", ["page"]),
-            ("跳转到{page}页面", ["page"]),
+            ("切换到{page}页面", ["page"]),
             ("访问{page}页面", ["page"]),
-            ("访问{page}模块", ["page"]),
+            ("切换{page}模块", ["page"]),
             ("访问{page}", ["page"]),
             ("请打开{page}模块", ["page"]),
             ("请打开{page}", ["page"]),
-            ("请打开{page}页面", ["page"]),
+            ("请切换到{page}页面", ["page"]),
             ("加载{page}模块", ["page"]),
-            ("加载{page}", ["page"]),
+            ("切换{page}", ["page"]),
             ("加载{page}页面", ["page"]),
         ]
     },