After buying my daughter a tablet with Google Assistant and watching how poor it was at giving her answers to her questions I thought I could do better. She’s 6 years old and asks 6 year old questions. “Why are firetrucks red?” to which it responds with some obscure mention of monty python and the queen. “What is a country?” which responds with jargon from wikipedia like “distinct geopolitical boundaries” which does not contain a single word she understands. I took a video of it because it was just so useless to a 6 year old that it was funny.
My solution: Take speech to text conversations from a huge range of 6 year olds. This becomes your corpus of knowable words for 6 year olds. Next, crawl the web and build your own index where you essentially penalise pages/text/sites that use words that are outside the corpus. What you end up with is the websites that are geared towards answering questions using wording that a 6 year old can understand. NOW if you ask google assistant as a 6 year old it’ll give you an answer that a 6 year old can understand.
The one caveat to this: I had not set my child’s age, so perhaps it is just that google didn’t know she was 6, and once I set that age it’ll give wildly different answers. I sort of hope it does.