WebSonar:The Speech Search Engine Based On Critical Information

The world has been becoming the age of information-economy. There are enormous information in the Internet, but it is very difficult to find the desired information. So search engine is going into people's eyes. Currently, the popular search engines are based on text information retrieval, like Google, Yahoo, Baidu and so on. For audio resource, they just retrieve the meta information, like author, title, description, file size and so on. They can't retrieve the content of the audio resource, like the content in a speech, the speaker in a speech. People will eagerly want to find the desired information according to the content of the audio resource because more and more audio resource will be placed in the Internet. 。 For this requirement, this paper proposed an audio search engine based on critical information. Critical information contains speakers and words in the audio resource. For example, we want to find some audio resource of "Bill Gates", including his talk records and other records referring him. The paper provided a prototype to search the critical information in the Internet. It integrated Keyword Spotting and Speaker Recognition into Search Engine. We can find the talk records for the specified person or the talk records which the specified person is referred.The main contributions of the work are the followings:1. Design a framework for Speech Search Engine2. Integrate Keyword Spotting and Speaker Recognition into Search Engine.3. Design an application based on Critical Persons.4. Design a partial optimized technology for Speech Search EngineThis work is supported by National Development & Innovation Committee (CNGI-04-12-2A), National Natural Science Foundation of P.R.China (60273059), National Science Fund for Distinguished Young Scholars (60525202), Program for New Century Excellent Talents in University (NCET-04-0545) and Key Program of Natural Science Foundation of China (60533040).