【Elasticsearch】全文检索 & 组合检索

1.全文检索

1.1 准备测试数据

创建一个索引。

PUT /products
{
  "mappings": {
    "properties": {
      "name": {"type": "text", "fields": {"keyword": {"type": "keyword"}}},
      "price": {"type": "double"},
      "category": {"type": "keyword"},
      "tags": {"type": "keyword"},
      "description": {"type": "text"},
      "stock": {"type": "integer"},
      "sku": {"type": "keyword"},
      "created_at": {"type": "date"},
      "metadata": {"type": "object"}
    }
  }
}

插入测试数据。

POST /products/_bulk
{"index":{}}
{"name":"Laptop X1","price":1299.99,"category":"electronics","tags":["new","sale"],"description":"High performance laptop","stock":50,"sku":"LP-X1-2023","created_at":"2023-01-15","metadata":{"weight":1.5,"color":"silver"}}
{"index":{}}
{"name":"Smartphone S10","price":899.99,"category":"electronics","tags":["new","popular"],"description":"Latest smartphone model","stock":120,"sku":"SP-S10-2023","created_at":"2023-02-20","metadata":{"weight":0.3,"color":"black"}}
{"index":{}}
{"name":"Wireless Headphones","price":199.99,"category":"audio","tags":["sale","popular"],"description":"Noise cancelling headphones","stock":75,"sku":"WH-200-2022","created_at":"2022-11-10","metadata":{"weight":0.25,"color":"white"}}
{"index":{}}
{"name":"Smart Watch","price":249.99,"category":"wearables","tags":["new","featured"],"description":"Fitness tracking smartwatch","stock":30,"sku":"SW-500-2023","created_at":"2023-03-05","metadata":{"weight":0.1,"color":"black"}}
{"index":{}}
{"name":"4K TV","price":1499.99,"category":"electronics","tags":["premium","large"],"description":"55-inch 4K television","stock":15,"sku":"TV-4K-55-2023","created_at":"2023-01-25","metadata":{"weight":18.5,"color":"black"}}
{"index":{}}
{"name":"Bluetooth Speaker","price":129.99,"category":"audio","tags":["portable"],"description":"Waterproof bluetooth speaker","stock":60,"sku":"BS-100-2022","created_at":"2022-12-15","metadata":{"weight":0.8,"color":"blue"}}
{"index":{}}
{"name":"Gaming Mouse","price":79.99,"category":"accessories","tags":["gaming"],"description":"High DPI gaming mouse","stock":90,"sku":"GM-X200","created_at":"2023-02-01","metadata":{"weight":0.12,"color":"rgb"}}
{"index":{}}
{"name":"External SSD 1TB","price":159.99,"category":"storage","tags":["fast","reliable"],"description":"Portable SSD drive","stock":45,"sku":"ESSD-1TB-2023","created_at":"2023-03-10","metadata":{"weight":0.05,"color":"gray"}}
{"index":{}}
{"name":"Keyboard Pro","price":109.99,"category":"accessories","tags":["ergonomic"],"description":"Mechanical keyboard","stock":25,"sku":"KB-PRO-2023","created_at":"2023-03-15","metadata":{"weight":1.1,"color":"black"}}
{"index":{}}
{"name":"Tablet T8","price":499.99,"category":"electronics","tags":["new","portable"],"description":"10-inch tablet","stock":40,"sku":"TAB-T8-2023","created_at":"2023-02-28","metadata":{"weight":0.5,"color":"silver"}}
{"index":{}}
{"name":"Camera DSLR","price":899.99,"category":"photography","tags":["professional"],"description":"24MP DSLR camera","stock":20,"sku":"CAM-DSLR-24","created_at":"2023-01-10","metadata":{"weight":0.7,"color":"black"}}
{"index":{}}
{"name":"Monitor 27\"","price":299.99,"category":"electronics","tags":["office"],"description":"27-inch office monitor","stock":35,"sku":"MON-27-2023","created_at":"2023-02-15","metadata":{"weight":4.2,"color":"black"}}

1.2 案例分析

1.2.1 match（分词检索）

对字段进行分词后匹配，支持模糊匹配和运算符。

GET /products/_search
{
  "query": {
    "match": {
      "description": {
        "query": "niose cancelling",  // 故意拼错 "noise" 测试模糊匹配
        "fuzziness": "AUTO"
      }
    }
  }
}

1.2.2 match_phrase（短语检索）

要求词语按顺序完整出现，可设置 slop，允许中间有其他词。

GET /products/_search
{
  "query": {
    "match_phrase": {
      "description": {
        "query": "high laptop",
        "slop": 1  // 允许中间有 1 个其他词
      }
    }
  }
}

1.2.3 match_phrase_prefix（短语前缀匹配）

短语匹配，但最后一个词支持前缀匹配。

GET /products/_search
{
  "query": {
    "match_phrase_prefix": {
      "name": {
        "query": "Smart Wa",  // 匹配 "Smart Watch" 等
        "max_expansions": 10  // 限制扩展数量
      }
    }
  }
}

1.2.4 multi_match（多字段匹配）

multi_match 检索适用于在多个字段上执行 match 检索的场景。它提供了一种方便的方法来在多个字段中间同时搜索指定的关键词，从而实现跨字段的高效检索。通过使用 multi_match 检索，用户可以简化复杂的多字段查询，优化搜索体验，并确保结果满足各种检索需求。

GET /products/_search
{
  "query": {
    "multi_match": {
      "query": "portable",
      "fields": ["name", "description", "tags"],
      "type": "best_fields"
    }
  }
}

由于涉及的字段不止一个，multi_match 检索在处理结果评分时采用特殊的评分机制，包括 most_fields、best_fields、cross_fields 等评分方式。这些评分方式确定了如何对每个字段获取的分数进行整合。

为了强调 tags 字段在搜索结果中的重要性，我们使用 ^3 来提高其权重。这意味着匹配 tags 字段的文档具有更高的相关性分数。

GET /products/_search
{
  "query": {
    "multi_match": {
      "query": "portable",
      "fields": ["name", "description", "tags^3"], 
      "type": "best_fields"
    }
  }
}

1.2.5 query_string（高级查询语法）

支持 Lucene 查询语法，功能强大但较复杂。

例如：查找在 name 或 description 字段中包含 laptop 或 smartphone，并且 price 字段值在 $100$ 到 $1000$ 之间的所有产品文档。

GET /products/_search
{
  "query": {
    "query_string": {
      "query": "(laptop OR smartphone) AND price:[100 TO 1000]",
      "fields": ["name", "description"],
      "default_operator": "AND"
    }
  }
}

1.2.6 simple_query_string

更简单的语法，对用户输入更友好，容错性更好。

例如，搜索同时满足以下条件的产品：

• 仅在商品名称（name）和描述（description）字段中搜索。
• 必须包含 speaker（由 +speaker 表示）。
• 必须不包含 blue（由 -blue 表示）。
• 最好包含 waterproof（没有前缀符号，作为可选条件）。

GET /products/_search
{
  "query": {
    "simple_query_string": {
      "query": "waterproof +speaker -blue", 
      "fields": ["name", "description"],
      "default_operator": "AND"
    }
  }
}

• + 必须包含，- 必须不包含。
• "default_operator": "AND" 表示当有多个搜索词时（没有 + / - 前缀的词），默认使用 AND 逻辑。
  • AND 操作符：提高精确度（结果更少但更相关）。
  • OR 操作符：提高召回率（结果更多但可能包含不相关项）。

如果用 SQL 表示，类似于：

SELECT * FROM products
WHERE (name LIKE '%speaker%' OR description LIKE '%speaker%')
AND (name NOT LIKE '%blue%' AND description NOT LIKE '%blue%')
AND (name LIKE '%waterproof%' OR description LIKE '%waterproof%')

注意：虽然 metadata.color 包含 blue，但没有检查 metadata.color 的内容，所以会按照上述内容返回。

如果真正目的是排除蓝色产品，应该这样查询：

GET /products/_search
{
  "query": {
    "bool": {
      "must": {
        "simple_query_string": {
          "query": "waterproof +speaker",
          "fields": ["name", "description"]
        }
      },
      "must_not": {
        "term": {
          "metadata.color": "blue"
        }
      }
    }
  }
}

1.3 对比总结表

2.组合检索

• must：查询结果必须满足指定条件。
• must_not：查询结果必须不满足指定条件。在此情况下，召回的数据评分为 $0$，且不考虑评分。
• filter：过滤条件，同样不考虑评分，召回的数据评分为 $0$。使用 filter 可以借助缓存机制提高查询性能。
• should：查询结果可以满足的部分条件，具体满足条件的最小数量由 minimum_should_match 参数控制。

🚀 Elasticsearch 查询语句中的 query 和 filter 具有不同的用途。

• query 用于评估文档相关性，并对结果进行评分，通常用于搜索场景。
• filter 用于筛选文档，不会对文档评分，通常用于过滤场景。

业务要求：查找符合以下条件的相关产品，其中：

• 必须在 category 或 description 中包含 electronics。
• 优先显示以下产品：
  • description 中提到 high performance 的产品。
  • 被标记为 popular 的产品。
  • 同时满足多个加分条件的产品会排名更靠前。

GET /products/_search
{
  "query": {
    "bool": {
      "must": [
        {
          "multi_match": {
            "query": "electronics",
            "fields": ["category^2", "description"],
            "type": "most_fields"
          }
        }
      ],
      "should": [
        {
          "match_phrase": {
            "description": {
              "query": "high performance",
              "slop": 2
            }
          }
        },
        {
          "match": {
            "tags": {
              "query": "popular"
            }
          }
        }
      ],
      "minimum_should_match": 1
    }
  }
}