ES实战-相关性搜索

wzerofeng2024-02-14 19:47

ES打分机制

1.TF-IDF 词频-逆文档频率

2.Okapi BM25

3.随机性分歧- DFR相似度

4.基于信息 - IB相似度

5.LM Dirichlet 相似度

6.LM Jelinek Mercer相似度
解释一个查询的结果集

bash 复制代码

curl -XPOST 'localhost:9200/get-together/_search?pretty' -H 'Content-Type: application/json' -d '
{
  "query": {
    "match": {
      "description": "elasticsearch"
    }
  },
  "explain": true
}'

使用function_score来定制得分

bash 复制代码

#在description字段中匹配到hadoop
#和logstash的文档,分数就乘以2或3
POST /get-together/_search
{
  "query": {
    "function_score": {
      "query": {
        "match": {
          "description": "elasticsearch"
        }
      },
      "functions": [
        {
          "weight": 2,
          "filter": {
            "term": {
              "description": "hadoop"
            }
          }
        },
        {
          "weight": 3,
          "filter": {
            "term": {
              "description": "logstash"
            }
          }
        }
      ]
    }
  }
}

合并得分

1.每个单独函数得分合并 score_mode

multiply,sum,avg,first,max,min

2.函数得分和原始得分合并 boost_mode

sum,avg,max,min,replace