Objective

Establish a large, on-line digital video library featuring fullcontent and knowledge-based search and retrieval.

Approach

Conduct research into the areas of speech recognition, image understanding and natural language processing to support the automatic preparation of diverse media for full-content and knowledge-based search and retrieval.

Accomplishments

Built testbed of over 1000 hours of hours of digital video, and audio, images, text and other related materials, consisting of science documentaries and lectures, and broadcast news content.

Automatic transcript generation for indexing.

Developed process for "video skimming," using language and image understanding techniques, that preserves video content while significantly reducing the amount of material to be reviewed.

Addressed issues pertinent to commercial partners, such as network billing, variable pricing and access control.

Explored performance issues related to Web-access and interoperability with other digital libraries.