So this is our grandiose idea to turn Open Mind Common Sense back into what it was meant to be: an ensemble method of loosely-linked AI techniques.
Thoughts so far:
- Write the core in Python 3 if possible. This prevents us from depending on too many modules... which I see as kind of a plus.
- Store information about the ensemble in MongoDB. - incremental CORONA over MongoDB to determine reliability
- No particular requirement that ensemble code should be in Python. It should just run, and do reasonable JSON I/O. - Probably over STOMP. With some approximation to JSON-RPC.