The image shows the underlying time-frequency characteristics of speech that are exploited by automatic recognition systems.

Human Language Technology

We apply machine-learning techniques to process data from speech, text, images, and multimedia. Human language researchers in our group have developed approaches for automatically identifying the words being spoken, the language being spoken, the topic of a conversation, and the speaker. Our cross-language retrieval systems could expedite analysts' search through foreign language documents. Recently, we have been developing social network analytics that are based on information extracted from speech, text, image, and video data combined with network communications and activities. These tools are helping law enforcement track down perpetrators engaging in illicit activity on the dark web.

Advancing Our Research

Featured Publications

Approaches for Language Identification in Mismatched Environments

Dec 13
Proceedings of SLT 2016, San Diego, Calif.

Detecting Depression using Vocal, Facial and Semantic Communication Cues(308.97 KB)

Oct 15
Proceedings of the Audio Visual Emotion Challenge and Workshop, Amsterdam, The Netherlands

Our Staff

View the biographies of members of the Human Language Technology Group.