Rough'n'Ready: Applying Speech and Language Technology to Real World Problems

Marie Meteer
BBN Technologies

The Rough'n'Ready audio indexing and browsing system brings together a set of leading edge technologies in speech recognition and language understanding in an innovative application for information access and knowledge management. This system provides a rough transcription that is ready for browsing, allowing a person to scan the information in an audio or video file, whether it is a current CNN news broadcast or a recording of a meeting, and search for particular topics, speakers, or mention of a name or place. One could ask to find where in the latest news reports that Bill Clinton talked about gun control and mentioned Littleton, Colorado, or one could ask to see a list of all of the topics that made it onto CNN Headline News this week and then select which to listen to.

In this talk I will provide an overview and demonstration of the Rough'n'Ready system and discuss the individual technologies that underlie it, including speech recognition, speaker demarcation and identification, topic segmentation and classification, and named-entity extraction.