multi_match.md

Multi Match Query

Another useful full text query is the multi match query. This query allows to search multiple fields at the same time. It provides a few more capabilities than a single match query. This query is a good start for more complex full text search.

Index Mapping

To illustrate search over multiple fields, first we create a new index with a mapping.

✅ Create a new index named multi_test with the following mapping

curl -X PUT 'http://localhost:9200/multi_test' -H 'Content-Type: application/json' -d '{
  "mappings": {
    "properties": {
      "title": {
        "type": "text",
        "analyzer": "standard"
      },
      "description": {
        "type": "text",
        "analyzer": "standard"
      },
      "tags": {
        "type": "keyword"
      }
    }
  }
}'

This creates a new index named multi_test and an index mapping with three properties

• title and description are using field type text using the standard analyzer
• tags with field type keyword.

Add Documents

Let's fill the empty index with some documents.

✅ Bulk upload documents to index multi_test

curl -H 'Content-Type: application/x-ndjson' -X POST 'http://localhost:9200/multi_test/_bulk' -d '
{"index":{"_index":"multi_test"}}
{"title": "Harry Potter And The Philosophers Stone", "description": "First movie of many", "tags": ["action", "kids", "magic"]}
{"index":{"_index":"multi_test"}}
{"title": "Star Trek", "description": "This is not Star Wars", "tags": ["space", "action", "better"]}
{"index":{"_index":"multi_test"}}
{"title": "Star Wars", "description": "Its the least favourite one", "tags": ["space", "action", "good"]}
{"index":{"_index":"multi_test"}}
{"title": "Star Lord", "description": "A tale of two wars", "tags": ["space", "action", "good"]}
{"index":{"_index":"multi_test"}}
{"title": "The Incredibles", "description": "A good animated spy movie", "tags": ["kids", "spy"]}
{"index":{"_index":"multi_test"}}
{"title": "Incredible Hulk", "description": "Hulk is the star", "tags": ["action", "hero"]}
{"index":{"_index":"multi_test"}}
{"title": "The Favourite", "description": "Incredible cast is incredible", "tags": ["costume", "play"]}
'

After these documents were added, they can be searched.

🔎 There are a few ways to check if an index exists. Elasticsearch offers cat API endpoints that return machine readable output instead of JSON. One such API is the cat indices API.

✅ Display the list of all indices

curl -X GET http://localhost:9200/_cat/indices?v

It lists a table with all indices, among the columns is the number of documents.

The basic structure of a multi_match query looks as follows:

{
  "query": {
    "multi_match": {
      "query": "<search-term>",
      "fields": ["title"]
    }
  }
}

The fields property takes a comma separated list of fields to search in. In the given example the list contains only field title. The <search-term> is the term to search for.

Exercise

Write multi_match queries to find a number of documents. Check the Elasticsearch documentation to learn what features you can use.

✅ Write a multi_match query to find documents whose text fields contain the term incredible

Possible Solution

curl -X POST 'http://localhost:9200/multi_test/_search?pretty' -H 'Content-Type: application/json' -d '{
  "query": {
    "multi_match": {
      "query": "incredible",
      "fields": ["title", "description"]
    }
  }
}'

✅ Write a multi_match query to find documents where the search term star wars is best found in one field.

Possible Solution

curl -X POST 'http://localhost:9200/multi_test/_search?pretty' -H 'Content-Type: application/json' -d '{
  "query": {
    "multi_match": {
      "query": "star wars",
      "fields": ["title", "description"],
      "type": "best_fields"
    }
  }
}'

The type field allows a few different values. Compare the results between queries using best_fields and most_fields, what is the difference?

✅ Write a multi_match query to rank matches in title higher than in description.

Possible Solution

curl -X POST 'http://localhost:9200/multi_test/_search?pretty' -H 'Content-Type: application/json' -d '{
  "query": {
    "multi_match": {
      "query": "incredible",
      "fields": ["title^3", "description"]
    }
  }
}'

This query boosts a match in title by a factor of 3 compared to a match in description.

Setting boost factors is a good way to rank documents according to the fields they match.