""" Phase 3 自动分类系统 基于商品名称和品牌信息进行智能分类 """ import re from typing import Dict, List, Tuple, Optional from dataclasses import dataclass from app.models.products import ProductCategory @dataclass class ClassificationRule: """分类规则""" category: ProductCategory keywords: List[str] brand_keywords: List[str] patterns: List[str] weight: float = 1.0 class ProductClassifier: """产品自动分类器""" def __init__(self): self.rules = self._load_classification_rules() self.compiled_patterns = self._compile_patterns() def _load_classification_rules(self) -> Dict[ProductCategory, ClassificationRule]: """加载分类规则""" rules = { # 美妆类 MC ProductCategory.MC: ClassificationRule( category=ProductCategory.MC, keywords=[ "口红", "唇膏", "唇釉", "唇彩", "唇蜜", "润唇膏", "唇线笔", "粉底", "粉底液", "气垫", "BB霜", "CC霜", "遮瑕", "散粉", "蜜粉", "眼影", "眼线", "眼线笔", "眼线液", "睫毛膏", "眉笔", "眉粉", "眉毛", "染眉膏", "腮红", "高光", "阴影", "修容", "化妆品", "彩妆", "美妆", "护肤品", "面霜", "面膜", "精华", "乳液", "爽肤水", "洁面", "卸妆", "防晒霜", "香水", "香氛", "指甲油", "美甲", "卸甲", "甲油胶" ], brand_keywords=[ "雅诗兰黛", "兰蔻", "迪奥", "香奈儿", "YSL", "圣罗兰", "阿玛尼", "娇兰", "SK-II", "资生堂", "FANCL", "POLA", "KOSE", "花王", "佳丽宝", "欧莱雅", "美宝莲", "卡姿兰", "完美日记", "花西子", "橘朵", "colorkey", "MAC", "NARS", "URBAN DECAY", "TOO FACED", "BENEFIT", "CLINIQUE" ], patterns=[ r".*[口唇][红膏釉彩蜜].*", r".*[粉底][液霜].*", r".*[眼影睫毛眉毛].*", r".*[面膜护肤化妆美妆].*", r".*[香水香氛].*" ], weight=1.0 ), # 服装类 AT ProductCategory.AT: ClassificationRule( category=ProductCategory.AT, keywords=[ "T恤", "衬衫", "衬衣", "polo", "POLO", "卫衣", "毛衣", "针织", "开衫", "外套", "夹克", "风衣", "大衣", "棉衣", "羽绒服", "马甲", "背心", "裤子", "牛仔裤", "休闲裤", "西裤", "短裤", "运动裤", "打底裤", "连体裤", "连衣裙", "半身裙", "长裙", "短裙", "A字裙", "包臀裙", "百褶裙", "西装", "套装", "正装", "礼服", "晚装", "婚纱", "旗袍", "睡衣", "家居服", "内衣", "文胸", "内裤", "袜子", "丝袜", "连裤袜", "上衣", "下装", "服装", "服饰", "童装", "男装", "女装" ], brand_keywords=[ "ZARA", "H&M", "UNIQLO", "优衣库", "GAP", "COS", "MUJI", "无印良品", "NIKE", "ADIDAS", "PUMA", "李宁", "安踏", "特步", "361", "匹克", "ONLY", "VERO MODA", "JACK JONES", "SELECTED", "太平鸟", "森马", "李维斯", "LEVIS", "CALVIN KLEIN", "TOMMY", "POLO", "拉夫劳伦" ], patterns=[ r".*[上衣下装外套裤子裙子].*", r".*[T恤衬衫卫衣毛衣].*", r".*[夹克风衣大衣羽绒服].*", r".*[连衣裙半身裙].*", r".*[牛仔休闲运动西装正装].*" ], weight=1.0 ), # 配饰类 AP ProductCategory.AP: ClassificationRule( category=ProductCategory.AP, keywords=[ "手表", "腕表", "智能手表", "电子表", "机械表", "石英表", "项链", "手链", "手镯", "戒指", "指环", "耳环", "耳钉", "耳坠", "胸针", "别针", "发夹", "发带", "头饰", "发圈", "头花", "帽子", "棒球帽", "鸭舌帽", "贝雷帽", "毛线帽", "遮阳帽", "渔夫帽", "围巾", "丝巾", "披肩", "头巾", "领带", "领结", "领巾", "皮带", "腰带", "吊带", "背带", "手套", "袖套", "太阳镜", "墨镜", "眼镜", "老花镜", "近视镜", "隐形眼镜", "配饰", "饰品", "首饰", "珠宝", "装饰品" ], brand_keywords=[ "劳力士", "欧米茄", "浪琴", "天梭", "卡西欧", "精工", "西铁城", "卡地亚", "蒂芙尼", "宝格丽", "梵克雅宝", "周大福", "周生生", "雷朋", "奥克利", "古驰", "普拉达", "路易威登", "爱马仕" ], patterns=[ r".*[手表腕表].*", r".*[项链手链戒指耳环].*", r".*[帽子围巾皮带].*", r".*[太阳镜墨镜眼镜].*", r".*[配饰饰品首饰].*" ], weight=1.0 ), # 箱包类 LA ProductCategory.LA: ClassificationRule( category=ProductCategory.LA, keywords=[ "包", "包包", "手提包", "单肩包", "双肩包", "斜挎包", "背包", "钱包", "钱夹", "卡包", "零钱包", "手拿包", "晚宴包", "手包", "公文包", "电脑包", "商务包", "书包", "学生包", "旅行包", "行李包", "拉杆箱", "行李箱", "旅行箱", "登机箱", "托运箱", "硬箱", "软箱", "化妆包", "洗漱包", "收纳包", "购物袋", "帆布袋", "环保袋", "腰包", "胸包", "挎包", "邮差包", "信封包", "贝壳包", "水桶包", "箱包", "皮具", "皮包", "真皮包", "帆布包" ], brand_keywords=[ "LV", "路易威登", "古驰", "GUCCI", "普拉达", "PRADA", "爱马仕", "HERMES", "香奈儿", "CHANEL", "迪奥", "DIOR", "芬迪", "FENDI", "巴黎世家", "新秀丽", "美旅", "外交官", "戴尔蒙", "卡拉羊", "爱华仕" ], patterns=[ r".*包.*", r".*[手提单肩双肩斜挎背]包.*", r".*[拉杆行李旅行登机]箱.*", r".*[钱包卡包零钱包].*", r".*[箱包皮具皮包].*" ], weight=1.0 ), # 鞋子类 SS ProductCategory.SS: ClassificationRule( category=ProductCategory.SS, keywords=[ "鞋", "鞋子", "运动鞋", "跑鞋", "篮球鞋", "足球鞋", "网球鞋", "板鞋", "休闲鞋", "帆布鞋", "小白鞋", "老爹鞋", "厚底鞋", "增高鞋", "皮鞋", "商务鞋", "正装鞋", "德比鞋", "牛津鞋", "布洛克鞋", "高跟鞋", "尖头鞋", "圆头鞋", "方头鞋", "细跟鞋", "粗跟鞋", "坡跟鞋", "平底鞋", "豆豆鞋", "乐福鞋", "玛丽珍鞋", "芭蕾鞋", "懒人鞋", "靴子", "短靴", "长靴", "马丁靴", "雪地靴", "雨靴", "工装靴", "凉鞋", "拖鞋", "洞洞鞋", "人字拖", "夹脚拖", "室内拖鞋", "童鞋", "婴儿鞋", "学步鞋" ], brand_keywords=[ "NIKE", "ADIDAS", "PUMA", "NEW BALANCE", "CONVERSE", "VANS", "李宁", "安踏", "特步", "361", "匹克", "乔丹", "鸿星尔克", "Dr.Martens", "马丁", "UGG", "Timberland", "添柏岚", "CHRISTIAN LOUBOUTIN", "Jimmy Choo", "Manolo Blahnik" ], patterns=[ r".*鞋.*", r".*[运动跑步篮球足球网球]鞋.*", r".*[休闲帆布皮鞋高跟平底]鞋.*", r".*[靴子短靴长靴雪地靴].*", r".*[凉鞋拖鞋].*" ], weight=1.0 ), # 内衣类 NY ProductCategory.NY: ClassificationRule( category=ProductCategory.NY, keywords=[ "内衣", "文胸", "胸罩", "bra", "BRA", "聚拢", "无钢圈", "调整型", "内裤", "三角裤", "平角裤", "丁字裤", "安全裤", "打底裤内穿", "睡衣", "睡袍", "睡裙", "家居服", "居家服", "loungewear", "保暖内衣", "秋衣", "秋裤", "长内衣", "长内裤", "打底衫内穿", "丝袜", "连裤袜", "长筒袜", "中筒袜", "短袜", "船袜", "隐形袜", "塑身衣", "塑形衣", "束腰", "美体", "收腹", "提臀", "哺乳", "孕妇", "产后", "哺乳内衣", "孕妇内衣" ], brand_keywords=[ "维多利亚的秘密", "VS", "Calvin Klein", "CK", "TRIUMPH", "黛安芬", "华歌尔", "爱慕", "古今", "曼妮芬", "都市丽人", "猫人", "南极人", "恒源祥", "俞兆林", "浪莎", "宝娜斯", "桑蚕丝" ], patterns=[ r".*[内衣文胸胸罩].*", r".*[内裤三角裤平角裤].*", r".*[睡衣睡袍家居服].*", r".*[保暖内衣秋衣秋裤].*", r".*[丝袜连裤袜长筒袜].*", r".*[塑身衣塑形衣束腰].*" ], weight=1.0 ) } return rules def _compile_patterns(self) -> Dict[ProductCategory, List]: """编译正则表达式模式""" compiled = {} for category, rule in self.rules.items(): compiled[category] = [re.compile(pattern, re.IGNORECASE) for pattern in rule.patterns] return compiled def classify_product( self, product_name: str, brand: Optional[str] = None, description: Optional[str] = None ) -> Tuple[ProductCategory, float]: """ 对产品进行分类 Args: product_name: 产品名称 brand: 品牌名称 description: 产品描述 Returns: Tuple[分类结果, 置信度] """ if not product_name: return ProductCategory.OTHER, 0.0 # 合并所有文本进行分析 text_to_analyze = product_name.lower() if brand: text_to_analyze += f" {brand.lower()}" if description: text_to_analyze += f" {description.lower()}" category_scores = {} for category, rule in self.rules.items(): score = 0.0 # 关键词匹配 keyword_matches = sum(1 for keyword in rule.keywords if keyword in text_to_analyze) score += keyword_matches * 2.0 # 品牌关键词匹配 if brand: brand_matches = sum(1 for brand_kw in rule.brand_keywords if brand_kw.lower() in text_to_analyze) score += brand_matches * 3.0 # 正则模式匹配 pattern_matches = sum(1 for pattern in self.compiled_patterns.get(category, []) if pattern.search(text_to_analyze)) score += pattern_matches * 1.5 # 应用权重 score *= rule.weight if score > 0: category_scores[category] = score if not category_scores: return ProductCategory.OTHER, 0.0 # 找到得分最高的分类 best_category = max(category_scores.keys(), key=lambda k: category_scores[k]) best_score = category_scores[best_category] # 计算置信度（标准化到0-1之间） total_score = sum(category_scores.values()) confidence = best_score / total_score if total_score > 0 else 0.0 confidence = min(confidence, 1.0) # 如果最高分太低，返回OTHER分类 if best_score < 1.0: return ProductCategory.OTHER, confidence * 0.5 return best_category, confidence def get_category_keywords(self, category: ProductCategory) -> List[str]: """获取指定分类的关键词""" rule = self.rules.get(category) return rule.keywords if rule else [] def get_all_categories(self) -> List[ProductCategory]: """获取所有可用分类""" return list(self.rules.keys()) def explain_classification( self, product_name: str, brand: Optional[str] = None, description: Optional[str] = None ) -> Dict: """ 解释分类结果 返回详细的分类分析信息 """ if not product_name: return {"error": "产品名称不能为空"} text_to_analyze = product_name.lower() if brand: text_to_analyze += f" {brand.lower()}" if description: text_to_analyze += f" {description.lower()}" analysis = { "input_text": text_to_analyze, "category_analysis": {}, "final_result": None } for category, rule in self.rules.items(): matched_keywords = [kw for kw in rule.keywords if kw in text_to_analyze] matched_brand_keywords = [kw for kw in rule.brand_keywords if kw.lower() in text_to_analyze] if brand else [] matched_patterns = [] for i, pattern in enumerate(self.compiled_patterns.get(category, [])): if pattern.search(text_to_analyze): matched_patterns.append(rule.patterns[i]) score = len(matched_keywords) * 2.0 + len(matched_brand_keywords) * 3.0 + len(matched_patterns) * 1.5 score *= rule.weight analysis["category_analysis"][category.value] = { "score": score, "matched_keywords": matched_keywords, "matched_brand_keywords": matched_brand_keywords, "matched_patterns": matched_patterns } # 获取最终分类结果 final_category, confidence = self.classify_product(product_name, brand, description) analysis["final_result"] = { "category": final_category.value, "confidence": confidence } return analysis