What we do
Language is a continuum of creation by the human mind. It is motivated by, and responsive to, human senses of touch, vision, hearing etc. and to scenarios created by their confluence. Our focus is to discover and understand the essence of human language through machine learning and signal processing techniques, as applied to the objects that create language scenarios. Our approach is a factor-and-unify one, wherein we view language as an information-rich signal produced by the human vocal tract system, and study separately how speech, text, images, sounds, touch and movement relate to language. We then try to figure out how we can bring this knowledge together in machine-understandable forms so that machines of the future can create and interpret language as well as humans can today.
Most of our work can be included under well-known areas of research: speech recognition, audio processing, image processing, artificial intelligence, semantics, statistics, signal processing, etc. However, there are subtle departures from the norm, which we don't currently attempt to categorize except as falling under the broad category of MLSP: Machine Learning for Signal Processing.
Some of our ongoing work is described on the projects page.