Pluggd’s “HearHere” Technology
Posted on 02/03/07 17:41
Tim Batzel and I met with Alex Castro, CEO of a local startup called Pluggd. We were particularly interested in learning about Pluggd’s “HearHere” technology which allows users to search for specific content within audio or video files. This allows you to jump to the topics you are most interested in hearing within a podcast and vidcast. What Google did for searching web documents, Pluggd is doing for audio and video.
Why is this relevant to educational technology? 73% of students reported listening to portions of the class podcasts multiple times in UW’s 2006 evaluation of current podcasting use. Imagine how much time would be saved if you could quickly jump to the portions of a class podcast that you needed to review. In contrast, imagine how arduous of a task it would be to sift through 10 weeks of podcasts to help you study for a final exam. Pluggd’s technology would have a very positive impact on the usability of educational podcasts – especially when we consider that many students use class podcasts to prepare for homework and exams, or to clarify concepts discussed in class.
Alex explained to us how their servers are constantly crawling the web and sifting through audio and video files to perform speech-to-text analysis. We are not talking about "Dragon NaturallySpeaking" here. We are talking about some serious department of defense- quality stuff running on Intel’s latest and greatest processing power. Another process tracks associations between words by looking at the results of the speech-to-text analysis and finding words that frequently appear in the same contexts. And if that weren’t enough, they also devour innumerable amounts of text web pages to build further word associations.
All this works to find related concepts appearing around the search terms that a user enters when trying to find content of interest within podcasts. The results are then displayed on the file’s timeline which is coded for relevance like a heat-map. The sections of the file that are related to the search term appear in red; click on red and you’d theoretically be listening to the details you missed on “elasticity of demand” in Professor Smith’s economics podcast.
I say theoretically because most of UW’s on-demand audio RSS feeds are hidden behind layers of security that prevent crawling and consumption by the general public.
Why is this relevant to educational technology? 73% of students reported listening to portions of the class podcasts multiple times in UW’s 2006 evaluation of current podcasting use. Imagine how much time would be saved if you could quickly jump to the portions of a class podcast that you needed to review. In contrast, imagine how arduous of a task it would be to sift through 10 weeks of podcasts to help you study for a final exam. Pluggd’s technology would have a very positive impact on the usability of educational podcasts – especially when we consider that many students use class podcasts to prepare for homework and exams, or to clarify concepts discussed in class.
Alex explained to us how their servers are constantly crawling the web and sifting through audio and video files to perform speech-to-text analysis. We are not talking about "Dragon NaturallySpeaking" here. We are talking about some serious department of defense- quality stuff running on Intel’s latest and greatest processing power. Another process tracks associations between words by looking at the results of the speech-to-text analysis and finding words that frequently appear in the same contexts. And if that weren’t enough, they also devour innumerable amounts of text web pages to build further word associations.
All this works to find related concepts appearing around the search terms that a user enters when trying to find content of interest within podcasts. The results are then displayed on the file’s timeline which is coded for relevance like a heat-map. The sections of the file that are related to the search term appear in red; click on red and you’d theoretically be listening to the details you missed on “elasticity of demand” in Professor Smith’s economics podcast.
I say theoretically because most of UW’s on-demand audio RSS feeds are hidden behind layers of security that prevent crawling and consumption by the general public.
