For some time, Google has been using neural matching to assist in delivering relevant web pages in response to queries in Google Search about 30% of the time. Now as of the November 2019 Local Search Update, Google has started using neural matching in Local Search.
As I understand it, neural matching as used by Google involves using artificial intelligence (AI) to match web pages to search queries using only the language of the query and the content of the web page, without reference to keywords or inbound links to the page.
But apparently Google's traditional ranking algorithm is used first, and pages are then re-ranked using neural matching.
This adds a new layer of complexity for anyone trying to do SEO for local search.
One possible approach would be to stuff pages with synonyms for the thing the page is about. But you can bet that Google would quickly get wise to that, and ranking would suffer.
Best approach to try might be to start with most important keyword for which you've optimized the page, do a Google search on that and take note of "Searches related to" that keyword, inject a few of the most relevant examples into the copy of the page where it makes sense to do so - like replacing duplicate mentions of the prime keyword.