Debarghya (Deedy) Das

This paper presents an articulatory biofeedback system and discusses new research methods made possible by this technology. The real-time electromagnetic articulography biofeedback system (RT-EMA) enables speakers to observe a visual representation of the movements of their speech articulators while they are speaking. Investigators can dynamically control the visual display of virtual targets or other objects in vocal tract space, track events involving interactions between virtual objects and… Show more

This paper presents an articulatory biofeedback system and discusses new research methods made possible by this technology. The real-time electromagnetic articulography biofeedback system (RT-EMA) enables speakers to observe a visual representation of the movements of their speech articulators while they are speaking. Investigators can dynamically control the visual display of virtual targets or other objects in vocal tract space, track events involving interactions between virtual objects and articulators, and define custom actions in response to such events. Preliminary findings from experimental studies and games employing biofeedback are reported, with emphasis on the potential applications of articulatory biofeedback for investigating questions of linguistic interest. Show less

We consider the problem of learning user preferences over robot trajectories in environments rich in objects and humans. This is challenging because the criterion defining a good trajectory varies with users, tasks and interactions in the environments. We use a cost function to represent how preferred the trajectory is; the robot uses this cost function to generate a trajectory in a new environment. In order to learn this cost function, we design a system - PlanIt, where non-expert users can… Show more

We consider the problem of learning user preferences over robot trajectories in environments rich in objects and humans. This is challenging because the criterion defining a good trajectory varies with users, tasks and interactions in the environments. We use a cost function to represent how preferred the trajectory is; the robot uses this cost function to generate a trajectory in a new environment. In order to learn this cost function, we design a system - PlanIt, where non-expert users can see robots motion for different asks and label segments of the video as good/bad/neutral. Using this weak, noisy labels, we learn the parameters of our model. Our model is a generative one, where the preferences are expressed as function of grounded object affordances. We test our approach on 112 different environments, and our extensive experiments show that we can learn meaningful preferences in the form of grounded planning affordances, and then use them to generate preferred trajectories in human environments.
Show less

Wrote a blog post about writing a MapReduce to scrape and statistically analyze the scores of some national-level exams in India, revealing critical anomalies. Over 400k views, and made front page of the Times of India, Hindustan Times, The Daily Mail, Scientific American, and more.