TED Talk Analysis

This research work captures the emotional trajectories or "styles of storytelling" in public speeches using computational analysis of the languaguistic styles. This analysis covers transcripts of over 2000 TED talks to estimate the potential audience responses. The ground truth of the analysis comes from over 5 million spontaneous annotations from the viewers in the TED Talk Website.

Learn more about this project


AutoManner is an intelligent interface for making public speakers aware of their mannerism. In this project we used an unsupervised algorithm to extract common repetitive patterns of body language, which was also developed in ROCHCI.

Learn more about this project

Automated Prediction of Job Interview

Ever wondered why you have been rejected from a job despite being a qualified candidate? In this project, we provide a computational framework to quantify human behavior in the context of job interviews. We build a model by analyzing 138 recorded interview videos (total duration of 10.5 hours) of 69 internship-seeking students from Massachusetts Institute of Technology (MIT) as they spoke with professional career counselors. Our framework is not only able to predict job interview performance, but also can generate automated recommendations.

Learn more about this project

LISSA - Live Interactive Social Skills Assistance

The LISSA system aims to help individuals improve their social skills by simulating face-to-face conversation and providing both real-time as well as post-session summary feedback on users’ nonverbal behavior. At the same time, it manages a fairly open-ended conversation with some degree of understanding. We have developed a HMM based model for generating real-time feedback.


ROC Comment

ROC Comment is an automated interface for generating natural language comments on behavioral videos. We focus on the domain of public speaking, which many people consider their greatest fear. We collect a dataset of 196 public speaking videos from 49 individuals and gather 12,173 comments, generated by more than 500 independent human judges. We then train a k-Nearest-Neighbor (k-NN) based model by extracting prosodic (e.g., volume) and facial (e.g., smiles) features. Given a new video, we extract features and select the closest comments using k-NN model. We further filter the comments by clustering them using DBScan, and eliminating the outliers. We are working on generating comments using deep learning. Our system is available online.

Visit the site


RocSpeak is a platform that allows you to record your speech on the browser and receive automated feedback on many aspects. In addition, you may share your video to receive personal feedback. Our algorithms will compile the personal feedback you have received with the automated features to give you helpful insight.

Visit the site