blocks|key|1600413|text|Elasticsearch中不同查询结果的组合通常是通过bool查询来实现的。可以使用function_score+query更改它们的组合方式。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|1600414|如果您需要组合不同的每个字段的评分函数(也称为similarity)，例如，使用BM25和DFR执行相同的查询并组合它们的结果，使用fields多次为同一字段建立索引可能会有所帮助。|1600415|现在让我来解释一下这个东西是如何工作的。|1600416|查找David+Gilmour的官方网站|1600417|假设我们有一个包含以下映射和示例文档的索引：|1600418|PUT+mysim
{
++"mappings":+{
++++"_doc":+{
++++++"properties":+{
++++++++"url":+{
++++++++++"type":+"keyword"
++++++++},
++++++++"title":+{
++++++++++"type":+"text"
++++++++},
++++++++"abstract":+{
++++++++++"type":+"text"
++++++++}
++++++}
++++}
++}
}

PUT+mysim/_doc/1
{
++"url":+"https://en.wikipedia.org/wiki/David_Bowie",
++"title":+"David+Bowie+-+Wikipedia",
++"abstract":+"David+Robert+Jones+(8+January+1947+–+10+January+2016),+known+professionally+as+David+Bowie+was+an+English+singer-songwriter+and+actor.+He+was+a+leading+..."
}

PUT+mysim/_doc/2
{
++"url":+"https://www.davidbowie.com/",
++"title":+"David+Bowie+%7C+The+official+website+of+David+Bowie+%7C+Out+Now+...",
++"abstract":+"David+Bowie+%7C+The+official+website+of+David+Bowie+%7C+Out+Now+Glastonbury+2000."
}

PUT+mysim/_doc/3
{
++"url":+"https://www.youtube.com/channel/UC8YgWcDKi1rLbQ1OtrOHeDw",
++"title":+"David+Bowie+-+YouTube",
++"abstract":+"This+is+the+official+David+Bowie+channel.+Features+official+music+videos+and+live+videos+from+throughout+David's+career,+including+Space+Oddity,+Changes,+Ash..."
}

PUT+mysim/_doc/4
{
++"url":+"www.davidgilmour.com/",
++"title":+"David+Gilmour+%7C+The+Voice+and+Guitar+of+Pink+Floyd+%7C+Official+Website",
++"abstract":+"David+Gilmour+is+a+guitarist+and+vocalist+with+British+rock+band+Pink+Floyd,+and+was+voted+No.+1+in+Fender's+Greatest+Players+poll+in+the+February+2006+Guitarist+..."
}|code-block|syntax|javascript|1600419|实际上，我们有一个David+Gilmour的官方网站，一个David+Bowie的官方网站，以及另外两个关于David+Bowie的页面。|1600420|让我们试着搜索一下David+Gilmour的官方网站：|1600421|POST+mysim/_search
{
++"query":+{
++++"match":+{
++++++"abstract":+"david+gilmour+official"
++++}++
++}
}|1600422|在我的机器上，这将返回以下结果：|1600423|++++"hits":+[
...
++++++++"_score":+1.111233,
++++++++"_source":+{
++++++++++"title":+"David+Bowie+%7C+The+official+website+of+David+Bowie+%7C+Out+Now+...",
...
++++++++"_score":+0.752356,
++++++++"_source":+{
++++++++++"title":+"David+Gilmour+%7C+The+Voice+and+Guitar+of+Pink+Floyd+%7C+Official+Website",
...
++++++++"_score":+0.68324494,
++++++++"_source":+{
++++++++++"title":+"David+Bowie+-+YouTube",
...|1600424|出于某种原因，David+Gilmour的页面并不是第一个。|1600425|如果我们从第一个查询中提取30%25的术语，就像最初的帖子所问的那样(让我们巧妙地选择gilmour使我们的示例更出色)，我们应该会看到改进：|1600426|POST+mysim/_search
{
++"query":+{
++++"match":+{
++++++"abstract":+"gilmour"
++++}++
++}
}|1600427|现在Elasticsearch只返回一个命中：|1600428|++++"hits":+[
...
++++++++"_score":+0.5956734,
++++++++"_source":+{
++++++++++"title":+"David+Gilmour+%7C+The+Voice+and+Guitar+of+Pink+Floyd+%7C+Official+Website",|1600429|比方说，我们不想放弃所有其他结果，只想重新排序，以便David+Gilmour的网站在结果中排名更高。我们能做什么？|1600430|使用简单的bool查询|1600431|bool查询的目的是以OR、AND或NOT的方式组合多个查询的结果。在我们的例子中，我们可以使用OR|1600432|POST+mysim/_search
{
++"query":+{
++++"bool":+{
++++++"should":+[
++++++++{
++++++++++"match":+{
++++++++++++"abstract":+"david+gilmour+official"
++++++++++}
++++++++},
++++++++{
++++++++++"match":+{
++++++++++++"abstract":+"gilmour"
++++++++++}
++++++++}
++++++]
++++}
++}
}|1600433|这似乎完成了工作(在我的机器上)：|1600434|++++"hits":+[
...
++++++++"_score":+1.3480294,
++++++++"_source":+{
++++++++++"title":+"David+Gilmour+%7C+The+Voice+and+Guitar+of+Pink+Floyd+%7C+Official+Website",
...
++++++++"_score":+1.111233,
++++++++"_source":+{
++++++++++"title":+"David+Bowie+%7C+The+official+website+of+David+Bowie+%7C+Out+Now+...",
...
++++++++"_score":+0.68324494,
++++++++"_source":+{
++++++++++"title":+"David+Bowie+-+YouTube",
...|1600435|bool查询在幕后所做的就是简单地对每个子查询的分数求和。在本例中，最高命中率的得分1.3480294是文档对上面两个独立查询的得分的总和：|1600436|>>>+0.752356+%2B+0.5956734
1.3480294000000002|1600437|但这可能还不够好。如果我们想将这些分数与不同的系数组合在一起呢？|1600438|组合具有不同系数的查询|1600439|为了实现这一点，我们可以使用function_score查询。|1600440|POST+mysim/_search
{
++"query":+{
++++"bool":+{
++++++"should":+[
++++++++{
++++++++++"function_score":+{
++++++++++++"query":+{
++++++++++++++"match":+{
++++++++++++++++"abstract":+"david+gilmour+official"
++++++++++++++}
++++++++++++},
++++++++++++"boost":+0.8
++++++++++}
++++++++},
++++++++{
++++++++++"function_score":+{
++++++++++++"query":+{
++++++++++++++"match":+{
++++++++++++++++"abstract":+"gilmour"
++++++++++++++}
++++++++++++},
++++++++++++"boost":+0.2
++++++++++}
++++++++}
++++++]
++++}
++}
}|1600441|在这里，我们用λ+=+0.8实现了原始post中的公式。|1600442|++++"hits":+[
...
++++++++"_score":+0.8889864,
++++++++"_source":+{
++++++++++"title":+"David+Bowie+%7C+The+official+website+of+David+Bowie+%7C+Out+Now+...",
...
++++++++"_score":+0.7210195,
++++++++"_source":+{
++++++++++"title":+"David+Gilmour+%7C+The+Voice+and+Guitar+of+Pink+Floyd+%7C+Official+Website",
...|1600443|在我的机器上，这仍然会产生“错误”的排序。|1600444|但是将λ更改为0.4似乎可以完成此任务！万岁！|1600445|如果我想结合不同的相似性呢？|1600446|如果您需要更深入，并能够修改Elasticsearch计算每个字段的相关性(称为similarity)的方式，可以通过定义custom+scoring+model来完成。|1600447|在我很难想象的情况下，你可能想要组合，比如说，BM25和DFR评分。Elasticsearch允许每个字段只有一个评分模型，但它也允许通过multi+fields对同一字段进行多次分析。|1600448|映射可能如下所示：|1600449|PUT+mysim
{
++"mappings":+{
++++"_doc":+{
++++++"properties":+{
++++++++"url":+{
++++++++++"type":+"keyword"
++++++++},
++++++++"title":+{
++++++++++"type":+"text"
++++++++},
++++++++"abstract":+{
++++++++++"type":+"text",
++++++++++"similarity":+"BM25",
++++++++++"fields":+{
++++++++++++"dfr":+{
++++++++++++++"type":+"text",
++++++++++++++"similarity":+"my_similarity"
++++++++++++}
++++++++++}
++++++++}
++++++}
++++}
++},
++"settings":+{
++++"index":+{
++++++"similarity":+{
++++++++"my_similarity":+{
++++++++++"type":+"DFR",
++++++++++"basic_model":+"g",
++++++++++"after_effect":+"l",
++++++++++"normalization":+"h2",
++++++++++"normalization.h2.c":+"3.0"
++++++++}
++++++}
++++}
++}
}|1600450|请注意，在这里我们定义了一个名为my_similarity的新相似度，它可以有效地计算DFR+(取自documentation)。|1600451|现在，我们将能够通过以下方式进行具有相似性组合的bool查询：|1600452|POST+mysim/_search
{
++"query":+{
++++"bool":+{
++++++"should":+[
++++++++{
++++++++++"match":+{
++++++++++++"abstract":+"david+gilmour+official"
++++++++++}
++++++++},
++++++++{
++++++++++"match":+{
++++++++++++"abstract.dfr":+"david+gilmour+official"
++++++++++}
++++++++}
++++++]
++++}
++}
}|1600453|请注意，我们对两个不同的字段执行相同的查询。这里abstract.dfr是一个“虚拟”字段，评分模型设置为DFR。|1600454|我还应该考虑什么？|1600455|在Elasticsearch中，分数是per-shard计算的，这可能会导致意外的结果。例如，IDF不是在整个索引上计算的，而是只在同一分片中的文档子集上计算的。|1600456|你可以阅读Here，Elasticsearch的主干是如何计算相关性分数的。|1600457|1600458|希望这能有所帮助！|1600459|entityMap|0|LINK|mutability|MUTABLE|url|https://www.elastic.co/guide/en/elasticsearch/reference/7.1/query-dsl-bool-query.html|1|https://www.elastic.co/guide/en/elasticsearch/reference/7.1/query-dsl-function-score-query.html|2|https://www.elastic.co/guide/en/elasticsearch/reference/7.1/similarity.html|3|https://www.elastic.co/guide/en/elasticsearch/reference/7.1/multi-fields.html|4|5|6|7|https://www.elastic.co/guide/en/elasticsearch/reference/7.1/index-modules-similarity.html|8|9|https://www.elastic.co/guide/en/elasticsearch/reference/current/index-modules-similarity.html#configuration|10|https://www.elastic.co/guide/en/elasticsearch/reference/7.1/consistent-scoring.html|11|https://www.elastic.co/guide/en/elasticsearch/guide/master/practical-scoring-function.html^0|S|4|17|E|S|4|0|17|E|1|0|14|4|19|3|1U|6|N|A|2|1U|6|3|0|0|0|0|0|0|0|0|0|0|0|15|7|0|0|0|0|0|5|4|5|4|4|0|0|4|B|2|E|3|I|3|1C|2|0|0|0|0|0|4|16|9|0|0|0|0|E|E|E|E|5|0|0|7|7|0|0|0|3|1|0|0|14|A|6|1P|K|7|0|N|4|S|3|1X|C|8|0|0|0|G|D|1E|D|9|0|O|4|0|0|O|C|0|0|J|9|A|0|5|4|B|0|0|0^^$0|@$1|2|3|4|5|6|7|3P|8|@$9|3Q|A|3R|B|C]|$9|3S|A|3T|B|C]]|D|@$9|3U|A|3V|1|3W]|$9|3X|A|3Y|1|3Z]]|E|$]]|$1|F|3|G|5|6|7|40|8|@$9|41|A|42|B|C]|$9|43|A|44|B|C]|$9|45|A|46|B|C]]|D|@$9|47|A|48|1|49]|$9|4A|A|4B|1|4C]]|E|$]]|$1|H|3|I|5|6|7|4D|8|@]|D|@]|E|$]]|$1|J|3|K|5|6|7|4E|8|@]|D|@]|E|$]]|$1|L|3|M|5|6|7|4F|8|@]|D|@]|E|$]]|$1|N|3|O|5|P|7|4G|8|@]|D|@]|E|$Q|R]]|$1|S|3|T|5|6|7|4H|8|@]|D|@]|E|$]]|$1|U|3|V|5|6|7|4I|8|@]|D|@]|E|$]]|$1|W|3|X|5|P|7|4J|8|@]|D|@]|E|$Q|R]]|$1|Y|3|Z|5|6|7|4K|8|@]|D|@]|E|$]]|$1|10|3|11|5|P|7|4L|8|@]|D|@]|E|$Q|R]]|$1|12|3|13|5|6|7|4M|8|@]|D|@]|E|$]]|$1|14|3|15|5|6|7|4N|8|@$9|4O|A|4P|B|C]]|D|@]|E|$]]|$1|16|3|17|5|P|7|4Q|8|@]|D|@]|E|$Q|R]]|$1|18|3|19|5|6|7|4R|8|@]|D|@]|E|$]]|$1|1A|3|1B|5|P|7|4S|8|@]|D|@]|E|$Q|R]]|$1|1C|3|1D|5|6|7|4T|8|@]|D|@]|E|$]]|$1|1E|3|1F|5|6|7|4U|8|@$9|4V|A|4W|B|C]]|D|@$9|4X|A|4Y|1|4Z]]|E|$]]|$1|1G|3|1H|5|6|7|50|8|@$9|51|A|52|B|C]|$9|53|A|54|B|C]|$9|55|A|56|B|C]|$9|57|A|58|B|C]|$9|59|A|5A|B|C]]|D|@]|E|$]]|$1|1I|3|1J|5|P|7|5B|8|@]|D|@]|E|$Q|R]]|$1|1K|3|1L|5|6|7|5C|8|@]|D|@]|E|$]]|$1|1M|3|1N|5|P|7|5D|8|@]|D|@]|E|$Q|R]]|$1|1O|3|1P|5|6|7|5E|8|@$9|5F|A|5G|B|C]|$9|5H|A|5I|B|C]]|D|@]|E|$]]|$1|1Q|3|1R|5|P|7|5J|8|@]|D|@]|E|$Q|R]]|$1|1S|3|1T|5|6|7|5K|8|@]|D|@]|E|$]]|$1|1U|3|1V|5|6|7|5L|8|@]|D|@]|E|$]]|$1|1W|3|1X|5|6|7|5M|8|@$9|5N|A|5O|B|C]]|D|@$9|5P|A|5Q|1|5R]]|E|$]]|$1|1Y|3|1Z|5|P|7|5S|8|@]|D|@]|E|$Q|R]]|$1|20|3|21|5|6|7|5T|8|@$9|5U|A|5V|B|C]]|D|@]|E|$]]|$1|22|3|23|5|P|7|5W|8|@]|D|@]|E|$Q|R]]|$1|24|3|25|5|6|7|5X|8|@]|D|@]|E|$]]|$1|26|3|27|5|6|7|5Y|8|@$9|5Z|A|60|B|C]]|D|@]|E|$]]|$1|28|3|29|5|6|7|61|8|@]|D|@]|E|$]]|$1|2A|3|2B|5|6|7|62|8|@]|D|@$9|63|A|64|1|65]|$9|66|A|67|1|68]]|E|$]]|$1|2C|3|2D|5|6|7|69|8|@$9|6A|A|6B|B|C]|$9|6C|A|6D|B|C]]|D|@$9|6E|A|6F|1|6G]]|E|$]]|$1|2E|3|2F|5|6|7|6H|8|@]|D|@]|E|$]]|$1|2G|3|2H|5|P|7|6I|8|@]|D|@]|E|$Q|R]]|$1|2I|3|2J|5|6|7|6J|8|@$9|6K|A|6L|B|C]]|D|@$9|6M|A|6N|1|6O]]|E|$]]|$1|2K|3|2L|5|6|7|6P|8|@$9|6Q|A|6R|B|C]]|D|@]|E|$]]|$1|2M|3|2N|5|P|7|6S|8|@]|D|@]|E|$Q|R]]|$1|2O|3|2P|5|6|7|6T|8|@$9|6U|A|6V|B|C]]|D|@]|E|$]]|$1|2Q|3|2R|5|6|7|6W|8|@]|D|@]|E|$]]|$1|2S|3|2T|5|6|7|6X|8|@]|D|@$9|6Y|A|6Z|1|70]]|E|$]]|$1|2U|3|2V|5|6|7|71|8|@]|D|@$9|72|A|73|1|74]]|E|$]]|$1|2W|3|-4|5|6|7|75|8|@]|D|@]|E|$]]|$1|2X|3|2Y|5|6|7|76|8|@]|D|@]|E|$]]|$1|2Z|3|-4|5|6|7|77|8|@]|D|@]|E|$]]]|30|$31|$5|32|33|34|E|$35|36]]|37|$5|32|33|34|E|$35|38]]|39|$5|32|33|34|E|$35|3A]]|3B|$5|32|33|34|E|$35|3C]]|3D|$5|32|33|34|E|$35|36]]|3E|$5|32|33|34|E|$35|38]]|3F|$5|32|33|34|E|$35|3A]]|3G|$5|32|33|34|E|$35|3H]]|3I|$5|32|33|34|E|$35|3C]]|3J|$5|32|33|34|E|$35|3K]]|3L|$5|32|33|34|E|$35|3M]]|3N|$5|32|33|34|E|$35|3O]]]]

Combination of results of different queries in Elasticsearch is commonly achieved with <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/query-dsl-bool-query.html" rel="nofollow noreferrer"><code>bool</code></a> query. Changes in the way they are combined can be made using <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/query-dsl-function-score-query.html" rel="nofollow noreferrer"><code>function_score</code></a> query.

In case you need to combine different per-field scoring functions (also known as <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/similarity.html" rel="nofollow noreferrer">similarity</a>), to, for instance, do the same query with <code>BM25</code> and <code>DFR</code> and combine their results, indexing the same field several times with use of <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/multi-fields.html" rel="nofollow noreferrer"><code>fields</code></a> can help.

Now let me explain how this thing works.

<h2>Find official website of David Gilmour</h2>

Let's imagine we have an index with following mapping and example documents:

<pre><code>PUT mysim
{
 "mappings": {
 "_doc": {
 "properties": {
 "url": {
 "type": "keyword"
 },
 "title": {
 "type": "text"
 },
 "abstract": {
 "type": "text"
 }
 }
 }
 }
}

PUT mysim/_doc/1
{
 "url": "https://en.wikipedia.org/wiki/David_Bowie",
 "title": "David Bowie - Wikipedia",
 "abstract": "David Robert Jones (8 January 1947 – 10 January 2016), known professionally as David Bowie was an English singer-songwriter and actor. He was a leading ..."
}

PUT mysim/_doc/2
{
 "url": "https://www.davidbowie.com/",
 "title": "David Bowie | The official website of David Bowie | Out Now ...",
 "abstract": "David Bowie | The official website of David Bowie | Out Now Glastonbury 2000."
}

PUT mysim/_doc/3
{
 "url": "https://www.youtube.com/channel/UC8YgWcDKi1rLbQ1OtrOHeDw",
 "title": "David Bowie - YouTube",
 "abstract": "This is the official David Bowie channel. Features official music videos and live videos from throughout David's career, including Space Oddity, Changes, Ash..."
}

PUT mysim/_doc/4
{
 "url": "www.davidgilmour.com/",
 "title": "David Gilmour | The Voice and Guitar of Pink Floyd | Official Website",
 "abstract": "David Gilmour is a guitarist and vocalist with British rock band Pink Floyd, and was voted No. 1 in Fender's Greatest Players poll in the February 2006 Guitarist ..."
}
</code></pre>

Practically speaking, we have an official website of David Gilmour, that one of David Bowie, and two other pages about David Bowie.

Let's try to search for David Gilmour's official website:

<pre><code>POST mysim/_search
{
 "query": {
 "match": {
 "abstract": "david gilmour official"
 } 
 }
}
</code></pre>

On my machine this returns the following results:

<pre><code> "hits": [
...
 "_score": 1.111233,
 "_source": {
 "title": "David Bowie | The official website of David Bowie | Out Now ...",
...
 "_score": 0.752356,
 "_source": {
 "title": "David Gilmour | The Voice and Guitar of Pink Floyd | Official Website",
...
 "_score": 0.68324494,
 "_source": {
 "title": "David Bowie - YouTube",
...
</code></pre>

For some reason, David Gilmour's page is not the first one.

If we take 30% of terms from the first query, like the original post is asking (let's cunningly select <code>gilmour</code> to make our example shine), we should see an improvement:

<pre><code>POST mysim/_search
{
 "query": {
 "match": {
 "abstract": "gilmour"
 } 
 }
}
</code></pre>

Now Elasticsearch only returns one hit:

<pre><code> "hits": [
...
 "_score": 0.5956734,
 "_source": {
 "title": "David Gilmour | The Voice and Guitar of Pink Floyd | Official Website",
</code></pre>

Let's say we don't want to discard all other results, just want to reorder so the David Gilmour's website is higher in the results. What can we do?

<h2>Use simple <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/query-dsl-bool-query.html" rel="nofollow noreferrer"><code>bool</code></a> query</h2>

The purpose of <code>bool</code> query is to combine results of several queries in <code>OR</code>, <code>AND</code> or <code>NOT</code> fashion. In our case we could go with <code>OR</code>:

<pre><code>POST mysim/_search
{
 "query": {
 "bool": {
 "should": [
 {
 "match": {
 "abstract": "david gilmour official"
 }
 },
 {
 "match": {
 "abstract": "gilmour"
 }
 }
 ]
 }
 }
}
</code></pre>

This seems to do the job (on my machine):

<pre><code> "hits": [
...
 "_score": 1.3480294,
 "_source": {
 "title": "David Gilmour | The Voice and Guitar of Pink Floyd | Official Website",
...
 "_score": 1.111233,
 "_source": {
 "title": "David Bowie | The official website of David Bowie | Out Now ...",
...
 "_score": 0.68324494,
 "_source": {
 "title": "David Bowie - YouTube",
...
</code></pre>

What <code>bool</code> query does under the hood is simply summing the scores per each subquery. In this case the top hit's score <code>1.3480294</code> is a sum of the document's score against two stand-alone queries we did above:

<pre><code>&gt;&gt;&gt; 0.752356 + 0.5956734
1.3480294000000002
</code></pre>

But this might not be good enough. What if we want to combine these scores with different coefficients?

<h2>Combine queries with different coefficients</h2>

To achieve this we can use <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/query-dsl-function-score-query.html" rel="nofollow noreferrer"><code>function_score</code></a> query.

<pre><code>POST mysim/_search
{
 "query": {
 "bool": {
 "should": [
 {
 "function_score": {
 "query": {
 "match": {
 "abstract": "david gilmour official"
 }
 },
 "boost": 0.8
 }
 },
 {
 "function_score": {
 "query": {
 "match": {
 "abstract": "gilmour"
 }
 },
 "boost": 0.2
 }
 }
 ]
 }
 }
}
</code></pre>

Here we implement the formula from the original post with <code>λ = 0.8</code>.

<pre><code> "hits": [
...
 "_score": 0.8889864,
 "_source": {
 "title": "David Bowie | The official website of David Bowie | Out Now ...",
...
 "_score": 0.7210195,
 "_source": {
 "title": "David Gilmour | The Voice and Guitar of Pink Floyd | Official Website",
...
</code></pre>

On my machine this still produces "wrong" ordering. 

But changing <code>λ</code> to 0.4 seems to do the job! Hooray!

<h2>What if I want to combine different similarities?</h2>

In case you need to go deeper, and be able to modify how Elasticsearch computes relevance per-field (which is called <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/similarity.html" rel="nofollow noreferrer">similarity</a>), it can be done via defining a <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/index-modules-similarity.html" rel="nofollow noreferrer">custom scoring model</a>.

In a case which I can hardly imagine, you may want to combine, say, <code>BM25</code> and <code>DFR</code> scoring. Elasticsearch only permits one scoring model per field, but it also allows to analyze the same field several times via <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/multi-fields.html" rel="nofollow noreferrer">multi fields</a>.

The mapping might look like this:

<pre><code>PUT mysim
{
 "mappings": {
 "_doc": {
 "properties": {
 "url": {
 "type": "keyword"
 },
 "title": {
 "type": "text"
 },
 "abstract": {
 "type": "text",
 "similarity": "BM25",
 "fields": {
 "dfr": {
 "type": "text",
 "similarity": "my_similarity"
 }
 }
 }
 }
 }
 },
 "settings": {
 "index": {
 "similarity": {
 "my_similarity": {
 "type": "DFR",
 "basic_model": "g",
 "after_effect": "l",
 "normalization": "h2",
 "normalization.h2.c": "3.0"
 }
 }
 }
 }
}
</code></pre>

Notice that here we defined a new similarity called <code>my_similarity</code> which effectively computes DFR (example taken from the <a href="https://www.elastic.co/guide/en/elasticsearch/reference/current/index-modules-similarity.html#configuration" rel="nofollow noreferrer">documentation</a>).

Now we will be able to do a <code>bool</code> query with a combination of similarities in the following way:

<pre><code>POST mysim/_search
{
 "query": {
 "bool": {
 "should": [
 {
 "match": {
 "abstract": "david gilmour official"
 }
 },
 {
 "match": {
 "abstract.dfr": "david gilmour official"
 }
 }
 ]
 }
 }
}
</code></pre>

Notice that we do the same query to two different fields. Here <code>abstract.dfr</code> is a "virtual" field with scoring model set to DFR.

<h2>What else should I consider?</h2>

In Elasticsearch scores are computed <a href="https://www.elastic.co/guide/en/elasticsearch/reference/7.1/consistent-scoring.html" rel="nofollow noreferrer">per-shard</a>, which can lead to unexpected results. For example, IDF is computed not on the whole index, but only on the subset of documents that are in the same shard.

<a href="https://www.elastic.co/guide/en/elasticsearch/guide/master/practical-scoring-function.html" rel="nofollow noreferrer">Here</a> you can read how Lucene, Elasticsearch's backbone, computes relevance scores.

<hr>

Hope that helps!

I am making a query A on elastic search and get the first 50 results. I also make a query B which contains the 30% of the terms of the query A. Each result of query A has a similarity score <code>scoreA</code> and each result of B has <code>scoreB</code>. 
What I am trying to achieve is combine the results of A and B to improve the Mean Average Precision of each imdividual query. One way that I found is to reorder the results based on this formula:

<pre><code>SIMnew = λ*scoreA + (1-λ)*scoreB
</code></pre>

where <code>λ</code> is a hyperparameter which I should tune. I noticed that this formula is very similar to Jelineck-Mercer smoothing which is implemented in Elastic Search (<a href="https://www.elastic.co/blog/language-models-in-elasticsearch" rel="nofollow noreferrer">https://www.elastic.co/blog/language-models-in-elasticsearch</a>).

Is there any default way to do this reordering with Elastic Search or the only way is a custom implementation?

(Given that I searched a lot about this formula and didn't find something usefull, it would be great if somenone gave me an intuition of how and why this works)

ElasticSearch-How to combine results of different queries to improve Mean Average Precision

我在elastic search上做了一个查询A，得到前50个结果。我还创建了一个查询B，其中包含查询A的30%的术语。查询A的每个结果都有一个相似度分数scoreA，并且B的每个结果都有scoreB。我试图实现的是将A和B的结果结合起来，以提高每个单独查询的平均精度。我发现的一种方法是根据这个公式对结果进行重新排序：...

问ElasticSearch-如何组合不同查询的结果以提高平均平均精度
EN

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问ElasticSearch-如何组合不同查询的结果以提高平均平均精度EN

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问ElasticSearch-如何组合不同查询的结果以提高平均平均精度
EN