Yep, you got it.

The ability to parse a human voice into text in near realtime?

Legitimate accomplishment.

Pretty high quality voice synthesis so it can talk back to you?

Also impressive!

But... you could have just plugged those things into... any other kind of system that evaluates inputs and generates outputs.

Could have just made it into basically an addon for wikipedia, webmd, whatever, with something like a very advanced version of a dialogue tree to basically frame the content, fucking clippy could do a reasonable job of summarizing and shortening long chunks of text into cliffs notes versions, 20 years ago.