I would train it off of all the unstructured text of my email, skype, jabber, salesforce, conversations. Maybe even throw in our dev wiki. I wanted to try doing this with the new IBM Watson API's. It was going to be amazing!
With a 2 month backpacking trip coming up, there was no time to build something like this. So, saner heads prevailed. I decided to write a simple web app that just gave canned deflecting answers, and just tell everyone that I wrote a super awesome system. So I wrote a page that just gives these answers randomly:
- "Yeah. I don't know."
- "Brian might know."
- "Object reference not set to instance of an object"
- "I'm busy."
- "Have you tried Googling it?"
- "What are you talking about?"
- "I don't remember..."
- "I'm in a meeting. I'll come grab you after.
I sent everyone this email:
Then gave everyone a link to the actual site: http://aaronmyster.github.io/laughing-shame/
It got a pretty good laugh. The interesting part was that people believed me! Not interesting in a "haha you're stupid for believing me" kind of way. More of a "It's crazy that this is possible" kind of way. It's amazing that we live in a time where something like this is even feasible.
Honestly, it's not a bad idea for a project...