Textbook queries video

As the amount of stored audio and video information burgeons, it is becoming increasingly clear that searching dynamic media needs to be as quick and easy as searching text.

Researchers from the University of Tsukuba in Japan, the Japan Science and Technology Corporation (CREST) and the Japanese National Institute of Advanced Industrial Science and Technology have found a simple way to improve audio and video searching.

The researchers' prototype uses lecture videos and related textbooks, and allows users to browse the textbooks to find information needed to query the system for related information on the videos. In the simplest example, a user could simply select a paragraph of text as a query.

The system automatically generates text of the audio tracks using speech recognition software, and uses uncorrected optical character recognition scans of related textbooks as a source for query information.

Even though neither technology is 100 percent accurate, the combination is more than accurate enough to find keywords, according to the researchers.

The method could be used to search any audio or video information source for which there is also related written material, including television news and newspaper articles, scientific lectures and papers, and TV cooking programs and recipes.

The work is one of many research projects aimed at making it easier to retrieve audio and video information. The method could be ready for practical use in three to six years, according to the researchers.