I'm sorry, but this is just patently wrong. However, it is an understandable misconception. The first speech synthesizer was invented in '36. This is much, much easier to create than a speech recognition machine. Speech synthesizing doesn't require much processing power, and would be doable in the 30's I would guess. Still very impressive though.
Speech recognition is a whole other beast, and requires alot of processing power and is waay beyond what was possible in the 30's, even command based recognition.
As I'm not just a naysayer I'll add my favorite from ww2 inventions:
Jerrycans! Which is, in some weird way, very awesome.