如何使用ES更有效率的进行多字段模糊匹配

今天药忘吃喽~ 2024-04-22 06:48 63阅读 0赞

### 背景 ###

有时候需要ES模糊多个多个和中文相关的字段，可以把多个字段合成一个逻辑意义上的字段进行模糊

### 相关信息 ###

此时需要两个配置:

1、copy\_to (将多个字段整合成一个字段)官方文档:https://www.elastic.co/guide/en/elasticsearch/reference/5.5/copy-to.html

2、ngram (分词器在对纯中文或者中英文混合相关等检索的时候很犀利，无脑的将词分隔成成为几个字连接起来)官方文档:https://www.elastic.co/guide/en/elasticsearch/reference/5.5/analysis-ngram-tokenizer.html

如下示配置看，将first\_name和last\_name字段copy到一个字段full\_name中

注意⚠️:(这个会不会产生一个情况?用户输入的前半段和后半段，命中了字段合在一起的中间部分，如A字段值:abc,B字段值为:def,把A、B字段合并后进行模糊分词匹配的时候会不会匹配中:cd这个值？)如果是使用copyto，不用担心这个。如果你使用的是binlog同步使用字符串拼接的方式产生的字段，会产生括号中所属的情况

PUT my_index
    {
        
      "mappings": {
        
        "my_type": {
        
          "properties": {
        
            "first_name": {
        
              "type": "text",
              "copy_to": "full_name" 
            },
            "last_name": {
        
              "type": "text",
              "copy_to": "full_name" 
            },
            "full_name": {
        
              "type": "text"
            }
          }
        }
      }
    }
    
    PUT my_index/my_type/1
    {
        
      "first_name": "John",
      "last_name": "Smith"
    }
    
    GET my_index/_search
    {
        
      "query": {
        
        "match": {
        
          "full_name": {
         
            "query": "John Smith",
            "operator": "and"
          }
        }
      }
    }

试一下没有进行额外配置ngrem默认的分词效果:会为字符创建一个offset，更加方便命中数据

GET _analyze
    {
        
      "tokenizer": "ngram",
      "text": "我的测试"
    }
    
    #产生的效果
    {
        
      "tokens": [
        {
        
          "token": "我",
          "start_offset": 0,
          "end_offset": 1,
          "type": "word",
          "position": 0
        },
        {
        
          "token": "我的",
          "start_offset": 0,
          "end_offset": 2,
          "type": "word",
          "position": 1
        },
        {
        
          "token": "的",
          "start_offset": 1,
          "end_offset": 2,
          "type": "word",
          "position": 2
        },
        {
        
          "token": "的测",
          "start_offset": 1,
          "end_offset": 3,
          "type": "word",
          "position": 3
        },
        {
        
          "token": "测",
          "start_offset": 2,
          "end_offset": 3,
          "type": "word",
          "position": 4
        },
        {
        
          "token": "测试",
          "start_offset": 2,
          "end_offset": 4,
          "type": "word",
          "position": 5
        },
        {
        
          "token": "试",
          "start_offset": 3,
          "end_offset": 4,
          "type": "word",
          "position": 6
        }
      ]
    }

进行ngrem配置

min\_gram和max\_gram 有助于分词的效率，类似一个窗口（如下配置像一个长度1～3单位的窗口，可以在字段字符上来回滑动分词），窗口长度拉的越长，匹配的更加具体，长度越小，匹配质量越低

PUT my_index
    {
        
      "settings": {
        
        "analysis": {
        
          "analyzer": {
        
            "my_analyzer": {
        
              "tokenizer": "my_tokenizer"
            }
          },
          "tokenizer": {
        
            "my_tokenizer": {
        
              "type": "ngram",
              "min_gram": 3,
              "max_gram": 3,
              "token_chars": [
                "letter",
                "digit"
              ]
            }
          }
        }
      }
    }
    
    POST my_index/_analyze
    {
        
      "analyzer": "my_analyzer",
      "text": "2 Quick Foxes."
    }

后续java示例伪代码

int from = (pageNum - 1) * pageSize;
            SearchSourceBuilder builder = new SearchSourceBuilder();
     		BoolQueryBuilder rootQuery = QueryBuilders.boolQuery();
    		BoolQueryBuilder boolQueryBuilder = QueryBuilders.boolQuery();
    		boolQueryBuilder.must(QueryBuilders.matchQuery("full_name", "模糊匹配的内容").operator(Operator.AND));
            rootQuery.filter(boolQueryBuilder);
            builder.setQuery(rootQuery);
            builder.from(from);
            builder.size(pageSize);
            SearchRequest searchRequest = new SearchRequest("indexName");
            searchRequest.source(builder);
            try {
        
                SearchResponse response = client.search(searchRequest);
                return response;
            } catch (IOException e) {
        
                throw new BaseException("es查询连接出错：" + e.getMessage());
            }catch (ElasticsearchException e){
        
                throw new BaseException("es查询出错：" + e.getMessage());
            }