何时在Elasticsearch中为Edge N Gram创建分析链

问题描述

我试图找出哪一种最适合增量搜索-edge n gramm或完成提示.到目前为止，我在Internet上阅读的内容表明，对于完成建议器，数据处理是在创建索引时完成的，而对于边缘n gram，它是在查询时完成的，因此它比完成建议器的处理速度慢.但是，就在Elasticsearch-definitve指南的那一刻，我读到对于边缘n gram来说，它也是在索引时完成的.现在我真的很困惑，任何人都可以清除，但是边缘n gram在内部起作用.

I was trying to figure out which one is better for incremental search - edge n gramm or completion suggester. So far what I have read on Internet suggested that for completion suggester processing of data is done at the time of index creation whereas for edge n gram its done at query time, hence its slower than completion suggester. However just now in the Elasticsearch - The definitve guide, book I read that for edge n gram also its done at indexing time. Now I am really confused, can anyone please clear however edge n gram internally works.

谢谢

推荐答案

两者均在索引时间执行，以建立专用的数据结构:

Both act at index time, building dedicated data structures:

N gram令牌生成器生成令牌:"hello world"变为"h"，"he"，"hel" ..."worl"，"world".使用通常的文本"(也称为字符串")映射类型.
• 完成建议程序生成一个图形:请参见 https://www.elastic.co/blog/you-complete-me .此时，有一个特殊的映射类型"completion".

• N gram tokenizer generates tokens: "hello world" becomes "h", "he", "hel"... "worl", "world". A usual "text" (aka "string") mapping type is used.
• completion suggester generates a graph: see https://www.elastic.co/blog/you-complete-me . At this point, there is a special mapping type "completion".

在搜索时，建议者便宜一些:

At search time, suggester are less expensive:

• N克令牌生成器:
  必须分析
  • 键入的文本，搜索结果词:搜索的"Hello Wor"应分析为"hello" +"wor"，然后搜索这两个词.
  • 但是，应将N gram标记器从分析中删除(在搜索和索引编制之间使用不同的分析器):搜索"Henry"将被分析为"h"，"he"，"hen"，"henr" ...会返回"hello"，因为它们共享相同的前缀"he".

  在两种情况下，您都可以使用自定义分析链(法语，德语，soundex ...):

  In both cases, you can use custom analysis chains (french, german, soundex...):

  N克:
  • 写时间:自定义分析+边缘ngram +键入文本"
  • 阅读时间:自定义分析+截断+搜索API
  • 写时间:自定义分析+输入完成"
  • 阅读时间:自定义分析+建议API

  这篇关于何时在Elasticsearch中为Edge N Gram创建分析链的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！