New!.. “Chat PG” - AI powered BiaB documentation using natural language queries - 08/14/23 03:59 PM
Here’s an alpha of
“Chat PG” - AI powered BiaB documentation using natural language queries
https://www.pgmusic.com/manuals/chat/
This uses a knowledge base of the online BiaB manual. And you can query it with natural language queries. Right now it just has the BiaB 2023 for windows essentials manual, and so it has no info about sales, or styles disks etc.
btw) It uses OpenAI “under the hood”, but is just accessed from our web page.
We can easily add lots to the knowledge base so that it can answer more questions. For example, we could “feed it” our FAQ and other knowledge documents.
We could also allow interested users to submit FAQ style question/answers or just knowledge points that they want added to the database. ‘For example, now it doesn’t know how to answer what XPro vs Xtra styles is, but it would when we tell it that.
It retains a log of questions that were asked, so we’d be able to see what questions it was missing to improve it by making sure the answer is in the documentation.
But for just having the essentials manual, it does a good job on most questions (maybe 70%), at least better than the current “search” we have that just uses Google to find web pages with key words. At least this answers a question or says “I don’t know”
“Chat PG” - AI powered BiaB documentation using natural language queries
https://www.pgmusic.com/manuals/chat/
This uses a knowledge base of the online BiaB manual. And you can query it with natural language queries. Right now it just has the BiaB 2023 for windows essentials manual, and so it has no info about sales, or styles disks etc.
btw) It uses OpenAI “under the hood”, but is just accessed from our web page.
We can easily add lots to the knowledge base so that it can answer more questions. For example, we could “feed it” our FAQ and other knowledge documents.
We could also allow interested users to submit FAQ style question/answers or just knowledge points that they want added to the database. ‘For example, now it doesn’t know how to answer what XPro vs Xtra styles is, but it would when we tell it that.
It retains a log of questions that were asked, so we’d be able to see what questions it was missing to improve it by making sure the answer is in the documentation.
But for just having the essentials manual, it does a good job on most questions (maybe 70%), at least better than the current “search” we have that just uses Google to find web pages with key words. At least this answers a question or says “I don’t know”