Elasticsearch13-relevance score相关性分数

问题：多分片下评分不准确，如何解决

因为计算评分都是在本地分片进行，并没有进行全局评分，就会造成误差较大。
目前大多解决方案是: 分片大小设置成一样的

 

 

 

 

 

 multi_match - best_fields让单个字段匹配多的排在前面

#想让单个字段匹配多的排在前面

GET product/_search
{
    "query": {
        "multi_match": {
           "query": "吃鸡手机",
           "fields": ["name", "desc"]
            # 默认   "type": "best_fields"
        }
    }
}

 

 

 

 

 

 

 

 

multi_match - most_fields多字段匹配

#使用多字段进行匹配<most_fields>
GET product/_search
{
    "query": {
        "multi_match": {
          "query": "超级快充",
          "fields": ["name", "desc"],
          "type": "most_fields"
        }
    }
}

 

 

 

 

 

 

 

 

 

multi_match - cross_fields

#cross_fields
#如果是and 多个字段加起来是'吴磊'
#比如 姓：吴    名：磊
GET teacher/_search
{
  "query": {
    "multi_match" : {
      "query":      "磊吴",
      "type":       "cross_fields",
      "fields":     [ "name.姓", "name.名" ],
      "operator":   "and"
    }
  }
}

　　

 

 

dis_max - tie_breaker 来调整TF 或IDF之间的比例，进而影响评分

GET product/_search
{
    "query": {
        "dis_max": {
            "queries": [
                {"match": {"name": "超级支持"}},
                {"match": {"desc": "超级支持"}}
            ],
            "tie_breaker": 0.7
        }
    }
}

 

 

 

查询修改score

GET product/_search
{
  "query": {
    "function_score": {
      "query": {
        "match_all": {}
      },
      "field_value_factor": {
        "field": "collected_num", #设置字段'collected_num'为_score分数
        "modifier": "log1p", # log(collected_num)
        "factor": 0.9 # log(collected_num)*0.9
      },
      "boost_mode": "multiply", log(collected_num)/0.9
      "max_boost": 3  # 最大分为3，超过3的都为3
    }
  }
}

 

 

 

 

排序位置前置(广告运营)

# 1.增加一个分数字段，该字段用来设置分数
# 2.给的钱越多，该值越大
# 3.关键字需要能匹配上
# 4.计算方式: query-score * script_score
GET product/_search
{
  "query": {
    "function_score": {
      "query": {
        "match_all": {}
      },
      "script_score": {
        "script": {
          "source": "Math.log(1 + doc['collected_num'].value)"
        }
      }
    }
  }
}