...for I have been playing around with a simulation that tries to learn the settings for the English stress system by searching through the incoming data in various ways. The search involves trying out different orderings of setting parameters, where the order matters since the model uses the previously set parameters to parse incoming data and decide what to do about the remaining unset parameters. I was rather afraid that it would turn out that no order would actually get a data-driven model to the right answer - but lo! I have determined today that a few such orderings do exist! The English stress system can be learnt from the data available to your average 1 year old! (The best ordering ends up being something like a breadth-first search of the parameter settings, with a delayed setting on one of them.)
And so I do a linguistic dance of great gleeful conquest! Hee!
:: dance dance ::