Any alternative solution to phrase query?

Elastic Stack Elasticsearch
1

I use Elasticsearch to store large amount of log data, so I'm particularly concerned abount storage costs.

This is a Lucene segment in my ES index:

1.5K   _3ee.fdm
2.4G   _3ee.fdt
172K   _3ee.fdx
3.4K   _3ee.fnm
172M   _3ee.kdd
518K   _3ee.kdi
555B   _3ee.kdm
720B   _3ee.si
279M   _3ee_Lucene80_0.dvd
2.0K   _3ee_Lucene80_0.dvm
390M   _3ee_Lucene84_0.doc
1.2G   _3ee_Lucene84_0.pos
452M   _3ee_Lucene84_0.tim
6.8M   _3ee_Lucene84_0.tip
2.3K   _3ee_Lucene84_0.tmd

It shows that the .pos file accounts for ~25% of the total size.
So I want to use index_options: docs to save the .pos file space and part of the .doc file space.

My question is: without the position data, is there any alternative solution to phrase query or LIKE "%xxx yyy%" in MYSQL?

For example, I have tried the script query, access _source field in the script and use contains to replace the phrase query:

GET my_index/_search
{
    "script_fields": {
        "source_field": {
            "script": {
                "source": "params['_source']['msg']"
            }            
        },
        "source_field_contains_xxx_yyy": {
            "script": {
                "source": "params['_source']['msg'].contains('xxx yyy')"
            }            
        }        
    }
}

But it seems that the _source field can only be accessed in the field context, can't be accessed in the filter context.