向量索引混合搜索

本文档介绍了 seekdb 全文索引和向量索引的混合搜索。

混合搜索（Hybrid Search）结合了基于向量的语义搜索和基于全文索引的关键词搜索，通过综合排序提供更准确、全面的搜索结果。向量搜索擅长语义近似匹配，但对精确的关键字、数字和专有名词等匹配能力较弱，而全文搜索能有效弥补这一不足。因此，混合搜索已成为向量数据库的关键特性之一，广泛应用于各类产品中。seekdb 通过整合其全文和向量索引能力，实现了高效的混合查询。

使用

混合搜索功能通过新增的系统包 DBMS_HYBRID_SEARCH 提供，其中包含 2 个子函数：

成员方法名称	功能介绍
`DBMS_HYBRID_SEARCH.SEARCH`	用于以 Json 格式返回搜索的结果，返回结果会按照相关性进行排序。
`DBMS_HYBRID_SEARCH.GET_SQL`	以字符串形式返回实际执行的 SQL 语句。

详细语法和参数说明请参见 DBMS_HYBRID_SEARCH。

使用场景说明

创建示例表并插入数据

为了演示混合搜索功能，本节将创建以下几张示例表并插入数据，这些表将用于下文不同场景的搜索示例。

products 表：一张基础商品信息表，用于演示普通标量搜索。它包含商品 ID、名称、描述、品牌、类别、标签、价格、库存、发布日期、是否在售以及一个向量字段 vec。

CREATE TABLE products (
  `product_id` varchar(50) DEFAULT NULL,
  `product_name` varchar(255) DEFAULT NULL,
  `description` text DEFAULT NULL,
  `brand` varchar(100) DEFAULT NULL,
  `category` varchar(100) DEFAULT NULL,
  `tags` varchar(255) DEFAULT NULL,
  `price` decimal(10,2) DEFAULT NULL,
  `stock_quantity` int(11) DEFAULT NULL,
  `release_date` datetime DEFAULT NULL,
  `is_on_sale` tinyint(1) DEFAULT NULL,
  `vec` VECTOR(4) DEFAULT NULL
);

插入数据。

INSERT INTO products VALUES
('prod-001', 'Gamer-Pro Mechanical Keyboard', 'A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.',
 'GamerZone', 'Gaming', 'best-seller,gaming-gear,rgb', 149.00, 100, '2023-07-20 00:00:00.000000', 1, '[0.5,0.1,0.6,0.9]'),

('prod-002', 'Gamer-Pro Headset', 'High-fidelity gaming headset with a noise-cancelling microphone.',
 'GamerZone', 'Gaming', 'best-seller,gaming-gear,audio', 149.00, 100, '2023-07-20 00:00:00.000000', 1, '[0.1,0.9,0.2,0]'),

('prod-003', 'Eco-Friendly Yoga Mat', 'A non-slip yoga mat made from sustainable and eco-friendly materials.',
 'NatureFirst', 'Sports', 'eco-friendly,health', 49.99, 200, '2023-04-22 00:00:00.000000', 0, '[0.1,0.9,0.3,0]');

products_fulltext 表：在 products 表的基础上，为 product_name、description 和 tags 列创建了全文索引，用于演示全文搜索。

CREATE TABLE products_fulltext (
    product_id VARCHAR(50),
    product_name VARCHAR(255),
    description TEXT,
    brand VARCHAR(100),
    category VARCHAR(100),
    tags VARCHAR(255),
    price DECIMAL(10, 2),
    stock_quantity INT,
    release_date DATETIME,
    is_on_sale TINYINT(1),
    vec vector(4),
    -- 在需要进行全文搜索的列上创建全文索引
    FULLTEXT INDEX idx_product_name(product_name),
    FULLTEXT INDEX idx_description(description),
    FULLTEXT INDEX idx_tags(tags)
);

插入数据。

INSERT INTO products_fulltext VALUES
('prod-001', 'Gamer-Pro Mechanical Keyboard', 'A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.',
 'GamerZone', 'Gaming', 'best-seller,gaming-gear,rgb', 149.00, 100, '2023-07-20 00:00:00.000000', 1, '[0.5,0.1,0.6,0.9]'),

('prod-002', 'Gamer-Pro Headset', 'High-fidelity gaming headset with a noise-cancelling microphone.',
 'GamerZone', 'Gaming', 'best-seller,gaming-gear,audio', 149.00, 100, '2023-07-20 00:00:00.000000', 1, '[0.1,0.9,0.2,0]'),

('prod-003', 'Eco-Friendly Yoga Mat', 'A non-slip yoga mat made from sustainable and eco-friendly materials.',
 'NatureFirst', 'Sports', 'eco-friendly,health', 49.99, 200, '2023-04-22 00:00:00.000000', 0, '[0.1,0.9,0.3,0]');

doc_table 表：一个包含标量列、向量列和全文索引列的文档表，用于演示全文搜索带标量过滤条件和混合搜索。

CREATE TABLE doc_table(
    c1 INT,
    vector VECTOR(3),
    query VARCHAR(255),
    content VARCHAR(255),
    VECTOR INDEX idx1(vector) WITH (distance=l2, type=hnsw, lib=vsag),
    FULLTEXT INDEX idx2(query),
    FULLTEXT INDEX idx3(content)
);

插入数据。

INSERT INTO doc_table VALUES
(1, '[1,2,3]', "hello world", "oceanbase Elasticsearch database"),

(2, '[1,2,1]', "hello world, what is your name", "oceanbase mysql database"),

(3, '[1,1,1]', "hello world, how are you", "oceanbase oracle database"),

(4, '[1,3,1]', "real world, where are you from", "postgres oracle database"),

(5, '[1,3,2]', "real world, how old are you", "redis oracle database"),

(6, '[2,1,1]', "hello world, where are you from", "starrocks oceanbase database");

products_vector 表：与 products 表结构类似，但明确为 vec 列创建了向量索引，用于演示纯向量搜索。

CREATE TABLE products_vector (
  `product_id` varchar(50) DEFAULT NULL,
  `product_name` varchar(255) DEFAULT NULL,
  `description` text DEFAULT NULL,
  `brand` varchar(100) DEFAULT NULL,
  `category` varchar(100) DEFAULT NULL,
  `tags` varchar(255) DEFAULT NULL,
  `price` decimal(10,2) DEFAULT NULL,
  `stock_quantity` int(11) DEFAULT NULL,
  `release_date` datetime DEFAULT NULL,
  `is_on_sale` tinyint(1) DEFAULT NULL,
  `vec` VECTOR(4) DEFAULT NULL,
  -- 在需要进行向量搜索的列上创建向量索引
  VECTOR INDEX idx1(vec) WITH (distance=l2, type=hnsw, lib=vsag)
);

插入数据。

INSERT INTO products_vector VALUES
('prod-001', 'Gamer-Pro Mechanical Keyboard', 'A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.',
 'GamerZone', 'Gaming', 'best-seller,gaming-gear,rgb', 149.00, 100, '2023-07-20 00:00:00.000000', 1, '[0.5,0.1,0.6,0.9]'),

('prod-002', 'Gamer-Pro Headset', 'High-fidelity gaming headset with a noise-cancelling microphone.',
 'GamerZone', 'Gaming', 'best-seller,gaming-gear,audio', 149.00, 100, '2023-07-20 00:00:00.000000', 1, '[0.1,0.9,0.2,0]'),

('prod-003', 'Eco-Friendly Yoga Mat', 'A non-slip yoga mat made from sustainable and eco-friendly materials.',
 'NatureFirst', 'Sports', 'eco-friendly,health', 49.99, 200, '2023-04-22 00:00:00.000000', 0, '[0.1,0.9,0.3,0]');

products_multi_vector 表：包含多个向量字段的表，用于演示多路向量搜索。

CREATE TABLE products_multi_vector (
    product_id VARCHAR(50),
    product_name VARCHAR(255),
    description TEXT,
    vec1 VECTOR(4),
    vec2 VECTOR(4),
    vec3 VECTOR(4),
    VECTOR INDEX idx1(vec1) WITH (distance=l2, type=hnsw, lib=vsag),
    VECTOR INDEX idx2(vec2) WITH (distance=l2, type=hnsw, lib=vsag),
    VECTOR INDEX idx3(vec3) WITH (distance=l2, type=hnsw, lib=vsag)
);

插入数据。

INSERT INTO products_multi_vector VALUES
('prod-001', 'Gamer-Pro Mechanical Keyboard', 'A responsive mechanical keyboard', '[0.5,0.1,0.6,0.9]', '[0.2,0.3,0.4,0.5]', '[0.1,0.2,0.3,0.4]'),
('prod-002', 'Gamer-Pro Headset', 'High-fidelity gaming headset', '[0.1,0.9,0.2,0]', '[0.3,0.4,0.5,0.6]', '[0.2,0.3,0.4,0.5]'),
('prod-003', 'Eco-Friendly Yoga Mat', 'A non-slip yoga mat', '[0.1,0.9,0.3,0]', '[0.4,0.5,0.6,0.7]', '[0.3,0.4,0.5,0.6]');

普通标量搜索

普通向量搜索的部分应用场景如下：

电商平台商品筛选：用户想要查看特定品牌的所有商品。例如，用户想要查看 GamerZone 品牌的所有商品。
内容管理系统：管理员需要筛选特定分类的文章或文档。例如，查找特定作者的所有文章。
用户管理系统：查找特定状态或角色的用户。例如，查找所有 VIP 用户。

示例如下：

设置搜索参数。

SET @parm = '{
"query": {
    "bool": {
        "must": [
        {"term": {"brand": "GamerZone"}}
        ]
    }
}
}';

搜索所有 brand 是 "GamerZone" 的记录。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products', @parm)); 

返回结果如下：

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products', @parm))                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "vec": "[0.5,0.1,0.6,0.9]",
    "tags": "best-seller,gaming-gear,rgb",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 1,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-004",
    "description": "A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.",
    "product_name": "Gamer-Pro Mechanical Keyboard",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
},
{
    "vec": "[0.1,0.9,0.2,0]",
    "tags": "best-seller,gaming-gear,audio",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 1,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-009",
    "description": "High-fidelity gaming headset with a noise-cancelling microphone.",
    "product_name": "Gamer-Pro Headset",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
}
] |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

普通标量的范围搜索

普通标量搜索的部分应用场景如下：

价格区间筛选：电商平台按价格范围筛选商品。例如，查找价格在 [30~80] 区间的商品。
时间范围查询：查找特定时间段内的订单或日志。例如，查找最近 30 天的订单。
数值范围过滤：筛选评分、库存数量等数值范围。例如，查找评分在 [4~5] 之间的商品。

示例如下：

设置搜索参数。

SET @parm = '{
"query": {
    "range" : {
    "price" : {
        "gte" : 30,
        "lte" : 80
    }
    }
}
}';

搜索所有 price 在 [30~80] 区间的记录。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products', @parm)); 

返回结果如下：

+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products', @parm))                                                                                                                                                                                                                                                                                                                                                                                          |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "vec": "[0.1,0.9,0.3,0]",
    "tags": "eco-friendly,health",
    "brand": "NatureFirst",
    "price": 49.99,
    "_score": true,
    "category": "Sports",
    "is_on_sale": 0,
    "product_id": "prod-003",
    "description": "A non-slip yoga mat made from sustainable and eco-friendly materials.",
    "product_name": "Eco-Friendly Yoga Mat",
    "release_date": "2023-04-22 00:00:00.000000",
    "stock_quantity": 200
}
] |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

全文搜索

全文搜索的部分应用场景如下：

文档搜索：在大量文档中搜索包含特定关键词的内容。例如，在 FAQ 中搜索包含 "如何使用" 的文档。
产品搜索：根据产品名称、描述进行模糊搜索。例如，搜索包含 "数据库" 的产品。
知识库搜索：在 FAQ、帮助文档中搜索相关问题。例如，在客服系统的知识库中搜索相关问题的答案。

示例如下：

设置搜索参数。

SET @query_str_with_mini = '{
"query": {
    "query_string": {
    "type": "best_fields",
    "fields": ["product_name^3", "description^2.5", "tags^1.5"],
    "query": "Gamer-Pro^2 keyboard^1.5 audio^1.2",
    "boost": 1.5
    }
}
}';

搜索 product_name、description 和 tags 字段中包含关键词 "Gamer-Pro"、"keyboard" 和 "audio" 的记录，并根据设置的字段和关键词权重进行排序。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_fulltext', @query_str_with_mini));

返回结果如下：

+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_fulltext', @query_str_with_mini))                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "vec": "[0.5,0.1,0.6,0.9]",
    "tags": "best-seller,gaming-gear,rgb",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 4.569735248749978,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-001",
    "description": "A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.",
    "product_name": "Gamer-Pro Mechanical Keyboard",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
},
{
    "vec": "[0.1,0.9,0.2,0]",
    "tags": "best-seller,gaming-gear,audio",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 1.7338881172399914,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-002",
    "description": "High-fidelity gaming headset with a noise-cancelling microphone.",
    "product_name": "Gamer-Pro Headset",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
}
] |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

全文搜索带标量过滤条件

全文搜索带标量过滤条件的部分应用场景如下：

精准搜索：在特定条件下进行文本搜索。例如，在已发布状态的文章中搜索特定关键词。
权限控制：在用户有权限的数据范围内进行搜索。例如，订单系统在特定时间段的订单中搜索商品信息。
分类搜索：在特定分类中进行关键词搜索。例如，用户系统在活跃用户中搜索特定用户信息。

示例如下：

设置搜索参数。

-- 过滤条件，指定标量过滤条件 c1 >= 2
SET @query_str = '{
"query": {
    "bool" : {
    "must" : [
        {"query_string": {
        "fields": ["query", "content"],
        "query": "hello what oceanbase mysql"}
        }
    ],
    "filter" : [
        {"range": {"c1": {"gte" : 2}}}  
    ]
    }
}
}';

搜索所有 c1 大于等于 2 的记录。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('doc_table', @query_str));

返回结果如下：

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('doc_table', @query_str))                                                                                                                                                                                                                                                                                                                                                                                                                                                                 |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "c1": 2,
    "query": "hello world, what is your name",
    "_score": 2.170969786679347,
    "vector": "[1,2,1]",
    "content": "oceanbase mysql database"
},
{
    "c1": 3,
    "query": "hello world, how are you",
    "_score": 0.3503184713375797,
    "vector": "[1,1,1]",
    "content": "oceanbase oracle database"
},
{
    "c1": 6,
    "query": "hello world, where are you from",
    "_score": 0.3503184713375797,
    "vector": "[2,1,1]",
    "content": "starrocks oceanbase database"
}
] |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

向量搜索

向量搜索的部分应用场景如下：

语义搜索：根据语义相似性查找相关内容。例如，在知识库中查找语义相关的问题和答案。
推荐系统：基于用户偏好推荐相似商品。例如，在电商平台上推荐相似商品。
图像搜索：通过图像特征查找相似图片。例如，在图片库中查找相似图片。
智能问答：在知识库中查找语义相关的问题和答案。例如，在客服系统的知识库中查找语义相关的问题和答案。

示例如下：

设置搜索参数。

-- field 指定向量字段，k 指定返回结果的数量（最近的 k 个结果），query_vector 指定查询的向量
SET @parm = '{
"knn" : {
    "field": "vec", 
    "k": 3, 
    "query_vector": [0.5,0.1,0.6,0.9] 
}
}';

搜索所有 vec 与 [0.5,0.1,0.6,0.9] 相似的记录。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_vector', @parm));

返回结果如下：

+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_vector', @parm))                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 |
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "vec": "[0.5,0.1,0.6,0.9]",
    "tags": "best-seller,gaming-gear,rgb",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 1.0,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-001",
    "description": "A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.",
    "product_name": "Gamer-Pro Mechanical Keyboard",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
},
{
    "vec": "[0.1,0.9,0.3,0]",
    "tags": "eco-friendly,health",
    "brand": "NatureFirst",
    "price": 49.99,
    "_score": 0.43405784,
    "category": "Sports",
    "is_on_sale": 0,
    "product_id": "prod-003",
    "description": "A non-slip yoga mat made from sustainable and eco-friendly materials.",
    "product_name": "Eco-Friendly Yoga Mat",
    "release_date": "2023-04-22 00:00:00.000000",
    "stock_quantity": 200
},
{
    "vec": "[0.1,0.9,0.2,0]",
    "tags": "best-seller,gaming-gear,audio",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 0.42910841,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-002",
    "description": "High-fidelity gaming headset with a noise-cancelling microphone.",
    "product_name": "Gamer-Pro Headset",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
}
] |
+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

向量搜索带标量过滤条件

向量搜索带标量过滤条件的部分应用场景如下：

精准搜索：在特定条件下进行文本搜索。例如，在已发布状态的文章中搜索特定关键词。
权限控制：在用户有权限的数据范围内进行搜索。例如，订单系统在特定时间段的订单中搜索商品信息。
分类搜索：在特定分类中进行关键词搜索。例如，用户系统在活跃用户中搜索特定用户信息。

示例如下：

设置搜索参数。

-- 指定标量过滤条件 brand = "GamerZone"
SET @parm = '{
"knn" : {
    "field": "vec",
    "k": 3,
    "query_vector": [0.1,0.5,0.3,0.7],
    "filter" : [
        {"term" : {"brand": "GamerZone"} }
    ]
}
}';

搜索所有 vec 与 [0.1,0.5,0.3,0.7] 相似的记录，并且 brand 是 "GamerZone" 的记录。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_vector', @parm));

返回结果如下：

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_vector', @parm))                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "vec": "[0.5,0.1,0.6,0.9]",
    "tags": "best-seller,gaming-gear,rgb",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 0.59850837,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-001",
    "description": "A responsive mechanical keyboard with customizable RGB lighting for the ultimate gaming experience.",
    "product_name": "Gamer-Pro Mechanical Keyboard",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
},
{
    "vec": "[0.1,0.9,0.2,0]",
    "tags": "best-seller,gaming-gear,audio",
    "brand": "GamerZone",
    "price": 149.00,
    "_score": 0.55175342,
    "category": "Gaming",
    "is_on_sale": 1,
    "product_id": "prod-002",
    "description": "High-fidelity gaming headset with a noise-cancelling microphone.",
    "product_name": "Gamer-Pro Headset",
    "release_date": "2023-07-20 00:00:00.000000",
    "stock_quantity": 100
}
] |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

多路向量搜索

多路向量搜索指在多个向量索引中进行搜索，并返回最相似的记录。

示例如下：

设置搜索参数。

-- 指定 3 路向量查询，每路查询指定向量索引字段，返回结果数量和查询向量
SET @param_multi_knn = '{
"knn" : [{
            "field": "vec1",
            "k": 5,
            "query_vector": [0.5,0.1,0.6,0.9]
        },
        {
            "field": "vec2",
            "k": 5,
            "query_vector": [0.2,0.3,0.4,0.5]
        },
        {
            "field": "vec3",
            "k": 5,
            "query_vector": [0.1,0.2,0.3,0.4]
        }
        ],
    "size" : 5
}';

执行查询并返回查询结果。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_multi_vector', @param_multi_knn));

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(DBMS_HYBRID_SEARCH.SEARCH('products_multi_vector', @param_multi_knn))                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "vec1": "[0.5,0.1,0.6,0.9]",
    "vec2": "[0.2,0.3,0.4,0.5]",
    "vec3": "[0.1,0.2,0.3,0.4]",
    "_score": 3.0,
    "product_id": "prod-001",
    "description": "A responsive mechanical keyboard",
    "product_name": "Gamer-Pro Mechanical Keyboard"
},
{
    "vec1": "[0.1,0.9,0.2,0]",
    "vec2": "[0.3,0.4,0.5,0.6]",
    "vec3": "[0.2,0.3,0.4,0.5]",
    "_score": 2.0957750699999997,
    "product_id": "prod-002",
    "description": "High-fidelity gaming headset",
    "product_name": "Gamer-Pro Headset"
},
{
    "vec1": "[0.1,0.9,0.3,0]",
    "vec2": "[0.4,0.5,0.6,0.7]",
    "vec3": "[0.3,0.4,0.5,0.6]",
    "_score": 1.86262927,
    "product_id": "prod-003",
    "description": "A non-slip yoga mat",
    "product_name": "Eco-Friendly Yoga Mat"
}
] |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set

全文与向量混合搜索

全文与向量混合搜索的部分应用场景如下：

智能搜索：结合关键词和语义理解的综合搜索。例如，用户输入 "我需要一个游戏键盘"，系统既匹配 "游戏"、"键盘" 关键词，又理解 "游戏设备" 的语义。
文档搜索：在大量文档中既支持精确关键词匹配，又支持语义理解。例如，搜索 "数据库优化"，既匹配包含这些词的文档，又找到关于 "性能调优"、"查询优化" 等语义相关的内容。
产品推荐：电商平台既支持商品名称搜索，又支持需求描述搜索。例如，根据用户描述 "适合办公的笔记本电脑"，既匹配关键词，又理解 "商务办公" 的语义需求。

示例如下：

设置搜索参数。

SET @parm = '{
  "query": {
    "bool": {
      "should": [
        {"match": {"query": "hi hello"}}, 
        {"match": { "content": "oceanbase mysql" }}
      ]
    }
  },
   "knn" : {
      "field": "vector",
      "k": 5,
      "query_vector": [1,2,3]
  },
  "_source" : ["query", "content", "_keyword_score", "_semantic_score"]
}';

执行查询并返回查询结果。

SELECT json_pretty(DBMS_HYBRID_SEARCH.SEARCH('doc_table', @parm)); 

返回结果如下：

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| json_pretty(dbms_hybrid_search.search('doc_table', @parm))                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| [
{
    "query": "hello world, what is your name",
    "_score": 2.835628417884166,
    "content": "oceanbase mysql database",
    "_keyword_score": 2.5022950878841663,
    "_semantic_score": 0.33333333
},
{
    "query": "hello world",
    "_score": 1.7219400929592013,
    "content": "oceanbase Elasticsearch database",
    "_keyword_score": 0.7219400929592014,
    "_semantic_score": 1.0
},
{
    "query": "hello world, how are you",
    "_score": 1.0096539326751595,
    "content": "oceanbase oracle database",
    "_keyword_score": 0.7006369426751594,
    "_semantic_score": 0.30901699
},
{
    "query": "real world, how old are you",
    "_score": 0.41421356,
    "content": "redis oracle database",
    "_keyword_score": null,
    "_semantic_score": 0.41421356
},
{
    "query": "real world, where are you from",
    "_score": 0.30901699,
    "content": "postgres oracle database",
    "_keyword_score": null,
    "_semantic_score": 0.30901699
}
] |

全文与向量 RRF 混合搜索

全文子查询和向量子查询的结果集默认采用加权混合。你可以通过 Rank 语法将融合方式配置为 RRF（Reciprocal Rank Fusion）排序混合。部分应用场景如下：

多维度排序：需要综合考虑多个搜索维度的结果。例如学术搜索系统，在论文库中搜索，既要考虑关键词匹配度，又要考虑语义相关性。
公平性要求：确保不同搜索方式的结果都能得到合理展示。例如，在电商平台上，既要考虑商品的标题、描述等文本信息，又要考虑商品的图片、视频等视觉信息。
复杂查询：涉及多个查询条件的复杂搜索场景。例如，医疗系统中，既要考虑患者的症状描述，又要考虑患者的病史、检查结果等。

示例如下：

设置搜索参数。

SET @rrf_query_param = '{
  "query": {
    "query_string": {
      "fields": ["title", "author", "description"],
      "query": "fiction American Dream"
    }
  },
  "knn" : {
    "field": "vector_embedding",
    "k": 5,
    "query_vector": [0.1, 0.2, 0.3, 0.4]
  },
  "rank" : {
    "rrf" : {
      "rank_window_size" : 10,
      "rank_constant" : 60
    }
  }
}';

RRF 算法通过融合多个子查询结果集的排名，计算最终的相关性分数。计算公式如下：

score = 0.0
for q in queries:
    if d in result(q):
        score += 1.0 / ( k + rank( result(q), d ) ) # K 常量是配置的 rank_constant
return score

总结

本文所提示例展示了混合搜索功能的强大应用价值：

智能搜索升级：在传统关键词搜索基础上融入语义理解，提供更精准、更符合用户意图的搜索结果。
优化用户体验：支持自然语言查询，简化操作，提升信息获取效率。
赋能多样业务：广泛应用于电商、内容管理、知识库、智能客服等场景，实现从基础筛选到智能推荐的全面覆盖。
融合技术优势：结合精确匹配与语义理解，显著提升搜索结果的准确性和全面性。

混合搜索功能是处理海量非结构化数据、构建智能搜索与推荐系统的理想选择。

向量索引混合搜索

使用

使用场景说明

创建示例表并插入数据

普通标量搜索

普通标量的范围搜索

全文搜索

全文搜索带标量过滤条件

向量搜索

向量搜索带标量过滤条件

多路向量搜索

全文与向量混合搜索

全文与向量 RRF 混合搜索

总结

相关文档

Contents

使用​

使用场景说明​

创建示例表并插入数据​

普通标量搜索​

普通标量的范围搜索​

全文搜索​

全文搜索带标量过滤条件​

向量搜索​

向量搜索带标量过滤条件​

多路向量搜索​

全文与向量混合搜索​

全文与向量 RRF 混合搜索​

总结​

相关文档​

Contents

使用

使用场景说明

创建示例表并插入数据

普通标量搜索

普通标量的范围搜索

全文搜索

全文搜索带标量过滤条件

向量搜索

向量搜索带标量过滤条件

多路向量搜索

全文与向量混合搜索

全文与向量 RRF 混合搜索

总结

相关文档