五金工业品商品名称解析与向量化技术方案

BRANDS = [
    "金杯", "舒友", "东成", "德力西", "正泰", "公牛", "西门子", "欧普",
    "雷士", "松下", "3M", "史丹利", "世达", "博世", "力丰", "宏发",
    "公牛", "鸿雁", "飞雕", "泰力", "龙胜", "飞利浦", "欧司朗",
]

SPEC_PATTERNS = [
    (r"(\d+\.?\d*)\s*[平方平米平]", "截面积", "mm²"),
    (r"(\d+\.?\d*)\s*[mM][mM]", "直径", "mm"),
    (r"(\d+\.?\d*)\s*A", "电流", "A"),
    (r"(\d+\.?\d*)\s*V", "电压", "V"),
    (r"(\d+\.?\d*)\s*W", "功率", "W"),
    (r"(\d+\.?\d*)\s*[Pp]", "功率", "P"),
    (r"(\d+\.?\d*)\s*寸", "尺寸", "寸"),
    (r"(\d+\.?\d*)\s*[Pp][Ss][iI]", "压力", "psi"),
    (r"(\d+\.?\d*)\s*[Hh][Zz]", "频率", "Hz"),
]

SPEC_NORMALIZE = {
    "1.5平": "1.5mm²", "1.5平方": "1.5mm²", "1.5平米": "1.5mm²",
    "2.5平": "2.5mm²", "2.5平方": "2.5mm²",
    "4平": "4mm²", "4平方": "4mm²",
    "6平": "6mm²", "6平方": "6mm²",
    "漏保": "漏电保护器", "空开": "空气开关",
    "排插": "插排", "插排": "插排",
}

# 搜索"金杯1.5平红色电线"
query = {
    "bool": {
        "should": [
            { "match": { "goods_name": "金杯 1.5平 红色 电线" }},
            { "match": { "pinyin": "jinbei dianxian" }},
            { "term": { "brand": "金杯" }},
            { "term": { "spec": "1.5mm²" }},
        ],
        "minimum_should_match": 2
    }
}

TRAIN_CONFIG = {
    "model": "hfl/chinese-macbert-base",
    "max_length": 64,
    "batch_size": 32,
    "learning_rate": 2e-5,
    "epochs": 10,
    "warmup_ratio": 0.1,
}

def encode_product_structured(product: dict) -> list[float]:
    vectors = []

    # 1. 品牌向量
    vectors.append(embed_text(f"品牌:{product['brand']}"))

    # 2. 品名向量
    vectors.append(embed_text(f"品名:{product['product_name']}"))

    # 3. 规格向量（归一化后）
    spec_normalized = normalize_spec(product['spec'])
    vectors.append(embed_text(f"规格:{spec_normalized}"))

    # 4. 品类向量
    vectors.append(embed_text(f"品类:{product['category']}"))

    # 5. 全量描述向量
    full_desc = f"{product['brand']} {product['product_name']} {product['spec']} {product['color']}"
    vectors.append(embed_text(full_desc))

    return concat_and_project(vectors, output_dim=256)

# 同商品不同表述 → 正例对
POSITIVE_PAIRS = [
    ("金杯1.5平方红色电线5卷", "红色1.5平方金杯电线5卷"),
    ("德力西63A漏电保护开关", "德力西漏电保护器63A2P"),
    ("舒友电工胶布红色1箱", "舒友红色胶布1箱（12卷）"),
]

# 不同商品 → 负例对
NEGATIVE_PAIRS = [
    ("金杯1.5平方红色电线", "金杯2.5平方红色电线"),
    ("德力西63A漏电开关", "德力西32A漏电开关"),
    ("东成20V锂电钻", "东成21V锂电钻套装"),
]

def contrastive_loss(anchor, positive, negative, margin=0.5):
    pos_sim = cosine_similarity(anchor, positive)
    neg_sim = cosine_similarity(anchor, negative)
    return max(0, neg_sim - pos_sim + margin)

CATEGORY_KEYWORDS = {
    "WIRE": ["电线", "电缆", "线缆", "BV", "BVR", "RVV"],
    "SWITCH": ["开关", "插座", "断路器", "漏电", "空开", "配电箱"],
    "LIGHT": ["灯泡", "灯管", "灯带", "吸顶灯", "射灯", "支架"],
    "TOOL": ["电钻", "冲击钻", "螺丝刀", "扳手", "钳子", "锯子", "锤子"],
    "FASTENER": ["螺丝", "螺母", "螺栓", "膨胀", "钉子", "铆钉"],
    "ADHESIVE": ["胶布", "胶带", "胶水", "ab胶", "防水胶带"],
    "PIPE": ["水管", "PPR", "PVC", "管件", "弯头", "三通", "阀门"],
    "BUILDING": ["水泥", "黄沙", "砖", "砂", "石子", "钢筋", "模板"],
}

AUGMENTATION_RULES = {
    "swap_order": {
        "input": "金杯电线",
        "outputs": ["电线金杯"]
    },
    "add_color": {
        "input": "金杯电线",
        "outputs": ["红色金杯电线", "金杯红色电线"]
    },
    "spec_variants": {
        "input": "1.5平方",
        "outputs": ["1.5平", "1.5mm²", "1.5平米"]
    },
    "brand_alias": {
        "input": "公牛",
        "outputs": ["GONGNIU", "Bull", "公牛电器"]
    },
    "category_expand": {
        "input": "漏保",
        "outputs": ["漏电保护器", "漏电开关", "RCD", "ELCB"]
    }
}

def auto_generate_training_data(product_template, count=10):
    """基于模板自动生成标注数据"""
    variations = []
    for _ in range(count):
        text = generate_variant(product_template, AUGMENTATION_RULES)
        labels = ner_model.predict_labels(text)  # 伪标签
        variations.append({"text": text, "labels": labels})
    return variations