Most of us are familiar with the speech recognition on today’s smartphones that basically is algorithms based on training data from human experts. MIT researchers’ are now taking the next step when developing a system that is unsupervised, which means it acts directly on raw speech files. The basic idea comes from that all languages has its own collection of phonemes, the English language has e.g. 35-45, from which spoken words are composed. With knowledge of a language phonemes it is easier for a machine to interpret speech. This is also similar to how children learn language, since they learn by finding patterns in the spoken language, not e.g. writing. Source: MITNews
Latest News: Artificial Intelligence – After Speech Recognition: What is next?
Bookmark the permalink.