Hi Steven,
That sounds like it would be challenging. While docx are parseable, there’s no real way to predict what the structure of the doc is and how it should be converted to a Swagger spec.
You can probably write something to do that for you, but unless your API is insanely big, it may be easier to just write the equivalent swagger spec manually.