Books are likely to be behind the SoTA

Developer conference talks can be a good source as an alternative, they tend to get into the nitty gritty, talk about the hard parts, and overcoming difficulties in production settings

I recall a talk from a Google i/o a few years back where they talked about how voice only is not the way, that depending on the question & answer, you want to move the user across modals

How we talk by Enfield was an interesting discussion on the mechanics and science of human to human speech. There’s some exploration of the nuances of speech in it.