Elasticsearch how to replace "terms" query? - elasticsearch

Question

Using Elasticsearch before version 6 the following query returned hits based on similarity to the query. Now in ES 6+ this returns hit based on whether they match any part of the query, all with the same score. This change breaks the rest of the more complicated query (not shown).
How can I get hit scores that match the older ES 5- for the following query
{
"query": {
"bool": {
"should": [
{
"terms": {
"some_field_name": [
"iPad Pro",
"iPhone 8"
]
}
}
]
}
}
}
In ES 5 this returns hits scored higher for matching all terms and lower for matching less. In ES 6+ this returns hits with only scores = 1 and so they are ranked with no regards to how many matched terms there were.
For instance a hit that has both terms is ranked higher in ES 5:
"_score": 0.87546873, when 2 of 4 terms match
"_score": 0.60353506, when 2 of 5 terms match
"_score": 0.13353139, when 1 of 4 terms match
This is much like the desired "cosine" similarity that we need (understood that this is not precisely how scores are created).
What query will return the same scores as ES 5 does for above query. In other words, what is the ES 6 equivalent query?

Answer 1

Note: Just posting it as a draft suggestion since discussion in comments is getting longer and clumsy.
UPDATE: Just checked both the methods I suggested and both seem to give similar scoring. Please try these methods and see if the scores you get are of any relevance to you.
I get that it doesn't work for terms but I suggested to replace it with either multiple "term" filter which also has a boost parameter or else replace it with "term_set" query and set required_matches param to "1" since we want the "or" behaviour.
1) Using terms_set query:
{
"query": {
"bool": {
"should": [
{
"terms_set": {
"some_field_name": {
"terms" : ["iPad Pro", "iPhone 8"],
"minimum_should_match_script": {
"source": "1"
}
}
}
}
]
}
}
}
2) Using multiple term filters instead of single terms filter:
{
"query": {
"bool": {
"should": [
{
"term": {
"some_field_name": "iPad Pro"
}
},
{
"term": {
"some_field_name": "iPhone 8"
}
}
]
}
}
}
If you were using boost in terms, then you should apply same boost in each of term blocks.