How should ElasticSearch be configured to index JSON meta data structures in text files?

Question

When adding front matter to a markdown document in i.e. GitLab, I am able to find the property using fulltext search.

---
title: "My subject"
author: "Jane Doe"
tags:
- tag: Book
- tag: Popular
- tag: Romance
---

# Chapter 1 - This is my Markdown

At first, the solution seemed a little bit odd.

If I need to find all Popular documents, I can search for "tag: popular" but not tag:popular. This isn't much of an issue if using YAML in the front matter section, but switching to JSON is a whole other beast.

;;;
{
  "title": "My subject",
  "author": "Jane Doe",
  "tags": [
    {
      "tag": "Book"
    },
    {
      "tag": "Popular"
    },
    {
      "tag": "Romance"
    }
  ]
}
;;;


# Chapter 1 - This is my Markdown

At first, the solution seemed a little bit odd.

Then I'd really like to only search using tag:popular instead of trying to figure out the free text syntax.

Note: Neither GitLab or Azure DevOps can presently render FrontMatter in any other format than YAML. So, to avoid bad rendering, JSON FrontMatter sections needs to be hidden using [comment]:# (hide this).

[JSON]:# (
;;;
{
  "title": "My subject",
  "author": "Jane Doe",
  "tags": [
    {
      "tag": "Book"
    },
    {
      "tag": "Popular"
    },
    {
      "tag": "Romance"
    }
  ]
}
;;;
)

# Chapter 1 - This is my Markdown

At first, the solution seemed a little bit odd.

Actually, the ElasticSearch engine i Azure DevOps do index datastructures found anywhere and make them searchable properties.

However, how should ElasticSearch be configured to support this feature on other vendors such as GitLab?