This morning I started working on a little project that I’ve been thinking about for several years. The goal is to be able to talk to the bot. That is, the IRC bot. Probably ‘rosie’, the bot on the TM3 channel. (No, the channel is not publically accessible, so if you’re not already on there, don’t bother asking.) The goal is to be able to speak to her, and have her respond. That is, verbally.
Most of the software is already out there to make this happen. I’ll be using Sphinx for the voice recognition, and Festival for the speech. I already have Festival working, and have had for many years. Often I’ll have Festival read IRC conversations to me while I’m doing something else. Sphinx, however, seems much harder. I will apparently need to train it to understand me. Although, by default, I’m amazed at how well it already understand me, albeit with a very very limited vocabulary.
I have not yet figured out how to get train it. The documentation is clearly written for developers, not for users. And I don’t yet know how to make it as convenient as Festival – ie, acting as a input/output pipe. But I just started this morning. So we’ll see.