You could use a look-up table, but that would not account for variations in tone and amplitude of spoken word, and the variety of inferences these variances imply.
Can you tell me how *you* recognize language? Most voice interface to this point is monotone and very limited vocabulary.