phpes短语精确搜索,Elasticsearch：在搜索引擎中如何实现完全匹配内容精确

phpes短语精确搜索,Elasticsearch：在搜索引擎中如何实现完全匹配（内容精确。。。本⽂地址：在有搜索引擎之前，我们查⽂档常使⽤顺序匹配。⽐如要搜索"我的祖国是花园"，需要在⽂档中顺序扫描，找到完全匹配的⼦句。在有了搜索引擎后，我们对查询语句做的处理就不⼀样了。我们通常会先分词，然后查找对应的词条索引，最后得到评分由⾼到低的⽂档列表。我⼀度以为没法实现完全匹配了，直到⼀个硬需求的出现。花了⼀天时间，把完全匹配⽤搜索引擎的思维整理出来。简要描述实现思路，字段按⼀字⼀词的形式分词，再利⽤短语查询来搜索。ES中，可以实现⼀字⼀词的的分词器是NGram。它其实是⼀个上下⽂相连续字符的分词⼯具，可以看官⽅⽂档中的例⼦。当我们将它min_gram和max_gram都设为1时，它会按⼀字⼀词的形式分词。⽐如“shinyke@189.cn”，分词的结果是["s","h","i","n","y","k","e","@","1","8","9",".","c","n"]。/index_name/{"settings":{"analysis":{"analyzer":{"charSplit":{"type":"custom","tokenizer":"ngram_tokenizer"}},"tokenizer":{"ngram_tokenizer":{"type":"nGram","min_gram":"1","max_gram":"1","token_chars":["letter","digit","punctuation"]}}}

}}以上语句中，构建了⼀个名为“charSplit”的分析器。它使⽤⼀个名为“ngram_tokenizer”的Ngram分词器。可以⽤如下语句测试charSplit分析器，可以看到⼀字⼀词的效果："测试语句"把这个分析器在mapping⾥⽤起来：..."sender":{"type":"string","store":"yes","analyzer":"charSplit","fields":{"raw":{"type":"string","index":"not_analyzed"}},...接下来就可以⽤match_phrase来实现完全匹配查询。/{index_name}/{type_name}/_search{"query":{"multi_match":{"query":"@189.cn","type":"phrase",//type指定为phrase"slop":0,//slop指定每个相邻词之间允许相隔多远。此处设置为0，以实现完全匹配。"fields":["sender"],"analyzer":"charSplit",//分析器指定为charSplit"max_expansions":1}},

"highlight":{//测试⾼亮是否正常"pre_tags":[""],"post_tags":[""],"fragment_size":100,"number_of_fragments":2,"require_field_match":true,"fields":{"sender":{}}}}phrase查询原始的作⽤是⽤来做短语查询，它有⼀个重要的特点：有顺序。我们利⽤了它匹配的有序性，限制slop为0，则可实现完全匹配查询。以上语句返回的结果是:{"took":18,"timed_out":false,"_shards":{"total":9,"successful":9,"failed":0},"hits":{"total":1,"max_score":0.40239456,"hits":[{"_index":"index_name","_type":"type_name","_id":"AU9OLIGOZN4dLecgyoKp","_score":0.40239456,

"_source":{"sender":lt189.cn>,李X<189.cn>,秦X<189.cn>,刘X<189.cn>"},"highlight":{"sender":[lt@189.cn>,李X<@189.cn>,秦纯X<@189.cn>,刘X<189773140"]}}]}}到此，就实现了完全匹配查询。实际环境中⽤NGram做⼀字⼀词