Brian Langner, Ph.D.


Postdoctoral Researcher
Language Technologies Institute
School of Computer Science
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15213




My primary research interests are in Speech Synthesis, Natural Language Generation, and Spoken Dialog (or Conversational) Systems. Specifically, my work has been in investigating methods of improving the quality of speech synthesis (and spoken language generation) so that it is clear and natural enough to be easily understood by users of speech systems – including populations that typically have difficulty understanding synthesized speech. This includes work on stylistic changes in spoken output as well as exploring improvements in presenting complex information using speech. Recent work has been in data-driven approaches to natural language generation, as well as investigating what makes spoken output understandable, and how to apply that to language generation for synthetic speech. These also lead towards general interests in Spoken Dialog to apply this work to interactive environments, as well as other areas of Computational Linguistics, Human-Computer Interaction, and Artificial Intelligence.

I previously worked on the Let's Go! project with my advisor, Dr. Alan W Black. This project involved building a spoken dialog system for bus information that can be readily used by non-native English speakers and the elderly, as well as the general population. Let's Go! and the underlying Olympus spoken dialog framework are now being used as a platform for the Dialog Research Center and the annual Spoken Dialog Challenge. I was also part of the original DialRC group.

I have additionally worked with Alex Hauptmann and Florian Metze on multimedia event detection, which involves multimedia analysis of consumer-grade video to detect and describe videos that contain specific predefined events. I was part of the speech and audio group, focusing on acoustic scene analysis (automatic semantic transcription of non-speech acoustic information) and event recounting (automatic generation of natural language video descriptions and explanations of detection decisions).

Other reasons to be at this page

Publications and Invited Talks

Valid XHTML 1.0! Valid CSS!