Index numbers in strings as numeric and words

[Michael_Lockwood]

Hi, I have a requirement where I want to search for numbers within a string by both the numeric value (e.g. 3) and the string value (e.g. "three").

For example given the following index configuration

PUT number-test
{
  "mappings": {
    "properties": {
      "text": {
        "type": "text", 
        "analyzer": "standard"
      }
    }
  }
}

PUT number-test/_doc/1
{
  "text": "this is the number three"
}

PUT number-test/_doc/2
{
  "text": "this is the number 3"
}

PUT number-test/_doc/3
{
  "text": "this is the number 2"
}

PUT number-test/_doc/4
{
  "text": "this is the number two"
}

Then the following search would return both documents 1 & 2.

POST number-test/_search
{
  "query": {
    "match": {
      "text": "three"
    }
  }
}

I know I could probably achieve this via synonyms. But I was hoping there was an automated way to do it (maybe through an additional tokenizer etc) that wouldn't require me having to configure every possible number as a synonym.

Thanks

[dadoonet (David Pilato)]

I don't think there is another solution than the synonyms or a Pattern replace.

May be using the new Machine Learning features for text would work though. But note that is part of the commercial license.