从 0 到 1 打造义乌购智能搜索接口：技术架构与性能优化全解析

import jieba
from transformers import BertTokenizer
# 加载行业词典与BERT分词器
jieba.load_userdict('yiwu_industry_dict.txt')
tokenizer = BertTokenizer.from_pretrained('bert-base-chinese')
def semantic_segment(keyword):
    """融合分词与语义扩展"""
    # 基础分词
    base_words = list(jieba.cut_for_search(keyword))
    # BERT语义扩展（提取核心词向量）
    bert_tokens = tokenizer.tokenize(keyword)
    # 合并去重
    return list(set(base_words + [t for t in bert_tokens if t not in ['[CLS]', '[SEP]']]))
# 测试：输入"时尚连衣裙" → 输出["时尚","连衣裙","潮流","女装"]

import pysolr
from rest_framework.views import APIView
from rest_framework.response import Response
# 连接SolrCloud集群（3主3从架构）
solr = pysolr.Solr(
    'http://solr-node1:8983/solr/yiwu_products',
    zkhost='zk-node1:2181,zk-node2:2181',  #  ZooKeeper集群
    timeout=5
)
class YiwuSearchAPI(APIView):
    def get(self, request):
        keyword = request.GET.get('keyword', '')
        if not keyword:
            return Response({"error": "关键字不能为空"}, status=400)
        # 语义分词+构建查询
        words = semantic_segment(keyword)
        query = " OR ".join([f"product_name:{w}^3 OR product_desc:{w} OR category:{w}^2" for w in words])
        try:
            # 执行搜索（按相关性+销量排序）
            results = solr.search(
                query,
                rows=20,
                sort='score desc, sales desc',
                fl='id,product_name,price,sales,shop_name'  # 只返回必要字段
            )
            return Response({
                "total": results.hits,
                "results": [dict(res) for res in results]
            })
        except Exception as e:
            return Response({"error": str(e)}, status=503)