context_suggestions.md

Context Suggestions

The last suggester described in this chapter is an extension of the suggest completion query type. It is a context based suggester that can be used to filter and/or boost suggestions by some criteria.

For example assume you want to suggest movies boosted by certain genres. In order to provide context based suggestions the completion field in the mapping is expanded with contexts information.

It's mandatory to provide a context when indexing / querying a context enabled completion field.

Index Mapping

To showcase how context based suggestions work, the following index mapping is used to set up a new index. See the Elasticsearch documentation on context suggesters.

✅ Create a new index named context_suggestions with the following mapping.

curl -X PUT 'http://localhost:9200/context_suggestions' -H 'Content-Type: application/json' -d '{
  "mappings": {
    "properties": {
      "title": {
        "type": "text",
        "analyzer": "standard",
        "fields": {
          "completion": {
            "type": "completion",
            "contexts": [
              {
                "name": "genre",
                "type": "category",
                "path": "genre"
              }
            ]
          }
        }
      },
      "genre": {
        "type": "keyword"
      },
      "year": {
        "type": "integer"
      }
    }
  }
}'

The mapping defines the following fields.

• title with field type text
• title.completion multi field with type completion with a context that point to field genre of the same document
• genre with field type keyword
• year with field type integer

The interesting part here is that the title.completion field that defines one category (genre) inside the contexts list.

Additional context mappings increase the index size for the completion field.

❗️ Only keyword or text fields can be parsed for context fields.

Add Documents

Once the index is created add a few documents to learn more about context suggestions.

✅ Bulk upload documents to index context_suggestions

curl -H 'Content-Type: application/x-ndjson' -X POST 'http://localhost:9200/context_suggestions/_bulk' -d '
{"index":{"_index":"context_suggestions"}}
{"title": "Star Trek Generations", "genre": "space action", "year": 1994}
{"index":{"_index":"context_suggestions"}}
{"title": "Star Wars A New Hope", "genre": "space action", "year": 1977}
{"index":{"_index":"context_suggestions"}}
{"title": "Stargate", "genre": "scifi", "year": 1994}
{"index":{"_index":"context_suggestions"}}
{"title": "Spiderman Homecoming", "genre": "superhero", "year": 2017}
{"index":{"_index":"context_suggestions"}}
{"title": "Spiderman Far From Home", "genre": "superhero", "year": 2019}
{"index":{"_index":"context_suggestions"}}
{"title": "Spiderman Into the Spiderverse", "genre": "animation", "year": 2018}
{"index":{"_index":"context_suggestions"}}
{"title": "Rogue One: A Star Wars Story", "genre": "space action", "year": 2016}
{"index":{"_index":"context_suggestions"}}
{"title": "A Star is Born", "genre": "biopic", "year": 2018}
{"index":{"_index":"context_suggestions"}}
{"title": "Star Trek Into Darkness", "genre": "action", "year": 2013}
{"index":{"_index":"context_suggestions"}}
{"title": "Third Star", "genre": "drama", "year": 2010}
{"index":{"_index":"context_suggestions"}}
{"title": "The Star Witness", "genre": "drama", "year": 1931}
'

Exercise & Example

A context suggestion query is more complex than the previous completion query. The query requires a contexts field in the suggest block.

✅ Run the following suggest query with search term "star" to get a list of results

curl -X POST 'http://localhost:9200/context_suggestions/_search?pretty' -H 'Content-Type: application/json' -d '{
  "suggest": {
    "text": "star",
    "context_completion_genre": {
      "completion": {
        "field": "title.completion",
        "contexts": {
          "genre": ["action"]
        }
      }
    }
  }
}'

Remember the completion field type indexes input text for a prefix based search, the text in title has to start with the given search term.

A context in the contexts field also supports boost factors. This provides a mean to rank values for a context differently. See the Elasticsearch documentation on context suggesters.

✅ Build a query that searches for "spider" in genres superhero and animation, that boosts animation by a factor of 2.0. Return the top 2 entries of this search.

Possible solution

Similar to the previous query there are now entries in the contexts field for genre. The entry for genre animation defines a boost factor of 2.0 which ranks these matches higher.

curl -X POST 'http://localhost:9200/context_suggestions/_search?pretty' -H 'Content-Type: application/json' -d '{
  "suggest": {
    "text": "spider",
    "context_completion_genre": {
      "completion": {
        "field": "title.completion",
        "size": 2,
        "contexts": {
          "genre": [
            { "context": "superhero" },
            { "context": "animation", "boost": 2.0 }
          ]
        }
      }
    }
  }
}'