Stemming/correction seems overly aggressive

I’ve noticed that search results seem to ve aggressively modifying query results. For example:

memory is matching Memorial
month is matching mother

It seems like a combination of typo correction and prefix matching. But it’s hard to diagnose what is causing something like this.

Without switching off features entirely, is there a way to tune these matches so I can filter out matches that are less relevant?

This is how my index is setup:

        "searchableAttributes": [...fields],
        "attributesForFaceting": ["filterOnly(user_id)"],
        "customRanking": ["desc(updated_at)"],
        "distinct": True,
        "attributeForDistinct": "id",
        "attributesToHighlight": [],
        "attributesToSnippet": ["*:20"],

This is how I’m performing a search (in python):
            "attributesToRetrieve": ["id"],
            "filters": f"user_id:{}",
            "hitsPerPage": 10,