I worked SO much this weekend, it was great. I made a test-standard data set for Ingush-Chechen-Batsbi, did a lot of refactoring and modularization of my alignment code, as well as setting up the Jenetics library so I can use genetic algorithms to tune my model parameters.
The results so far are fantastic but i have a
lot of work still ahead. One rather interesting result so far is using a gap penalty (at least, a constant one) is actually bad:
Admittedly, I have not yet tried this with my Indo-Iranian data. I had been using a penalty of 6, which actually was a very bad choice.