tag:sindro.me,2012:/2009/4/21/a-tweeting-geeky-parrot/comments Mephisto Drax 2009-04-24T18:12:22Z Antonio tag:sindro.me,2009-04-21:130:131 2009-04-24T15:52:20Z 2009-04-24T15:52:20Z <p>Maybe you could also build a live feedback system to make its learning faster. When he says something, you post it on twitter, and a software chooses the most similar real word from a dictionary, then synthesize it in waveform and play through an audio speaker near him. If he speaks another time in a defined limited time interval, the same process happens and a servo motor opens the way for a sweet something to fall in his dish, ready to be eat. The amount of sweet stuff falling is proportional to the similarity of the word to the previous one. When the twitted word is 100% existing, the software forces an audio of that word followed by another one, to stimulate learning pronuciation of a new real word. So he’s learning A+B; than it will learn B+C, then B+D, etc. When the software estimates that he knows enough words to speak a meaningful sentence (i.e. something at least 5 words long that searched on Google with quotes returns a good amount of results), the process goes like the single word one, but this time with the full meaningful sentence. The posted tweets will show the learning curve of the parrot to learn that sentence :)</p>