Filters with dynamic fields in elasticsearch

Question

I am switching a very old version of ElasticSearch to version 6.5.

...  
  "text_mined_entities": {
            "nlp": {
                "abbreviations": [],
                "chunks": [],
                "recurring_chunks": [],
                "tagged_entities_grouped": {
                    "NEURO|SCICRUNCH": [
                        {
                            "category": "NEURO",
                            "end": 41,
                            "label": "Infant",
                            "match": "infant",
                            "original_value": "Infant",
                            "reference": "BIRNLEX695",
                            "reference_db": "SCICRUNCH",
                            "sentence": 0,
                            "start": 35
                        },
...

I am wanting to filter on the text_mined_entities.nlp.tagged_entities_grouped.*.reference fields ( which are stored as 'keyword' ), but haven't had much luck. Something like:

GET _search
{
  "query": {
    "bool": {
      "filter": { "term": {
        "text_mined_entities.nlp.tagged_entities_grouped.*.reference": "BIRNLEX695"

      }}

    }
  }
}

Any suggestions? Thanks.

Answer 1

Wildcard on fields can't be applied on term query. Instead you can use query_string which supports wildcard on field as well. So following will work:

Assuming text_mined_entities and nlp are of type nested

{
  "query": {
    "nested": {
      "path": "text_mined_entities.nlp",
      "query": {
        "query_string": {
          "query": "BIRNLEX695",
          "fields": [
            "text_mined_entities.nlp.tagged_entities_grouped.*.reference"
          ]
        }
      }
    }
  }
}

Update (if text_mined_entities & nlp are object type and not nested):

{
  "query": {
    "query_string": {
      "query": "BIRNLEX695",
      "fields": [
        "text_mined_entities.nlp.tagged_entities_grouped.*.reference"
      ]
    }
  }
}