Death Valley Dunes©Gary Hayes 2005The world is getting ready to personalize its media. Beyond simply attaching a bunch of keywords to your blog or website so the google bots can make you exist, we are now entering a world where your audio and video files can be trawled too – images and sound analysed and rich descriptive metadata added and embedded, ready for th 21st century. I have made a few posts in this space over the last few months, Pieces of our Stories, The MyEdit Phenomenon, Dawn of Video Personalization and many others so it is great that there are very current mass market tools appearing that will make sound (or as it is now called Podcasts 😉 as easy to search as text. Once this metadata is added linear sequences can be split up and delivered as cross media experiences amongst other things. It would be easy to take a 2 hour vlog and deliver the first segment as video on the web, the second as a text installment for your phone and others as podcasts or a narrow band web page with segment links and so on.

The wired article Podcast Chaos Be Gone talks about Podzinger, Podscope and blinkx, two engines that scour your audio files and create a metadata index based on a range of sophisticated speech to text recognition engines. Once we have a time-stamped index then it

lets users jump to the spot in a podcast where their search term appears, rather than forcing them to scan an entire program for pertinent parts as blinkx does. Its minimalist design has an uncluttered search page, à la Google. And Podcasters can link to a searchable index of their content in order to sell sponsored links to text in the index provided by Podzinger.(snip) But Podzinger’s information is more extensive and the results include a counter indicating where in the podcast the snippet is located. And Podzinger allows users to click on the Play button to hear the excerpt and determine its relevancy before clicking on a link to download the entire podcast.

It is not new to many people that real time engines exist that convert clean speech into text. The dictation engines, ViaVoice and Dragon NaturallySpeaking etc have been around for a while and do a great job as long as you have a relatively noise free environment, so I wonder how good Podzinger will be? Another clever sound analyser Podscope which has been around for 7 years is being eye’d by Google

A third search engine, launched last spring by TVEyes and called Podscope, searches podcasts but scans only for the sounds of syllables rather than full words. AOL recently announced plans to integrate Podscope into the portal’s search page.

Of course as long as video content has a good clean audio track – which is certainly more likely in professional video content (the sort of stuff we used to get on that old TV thing) then indexing video also gets into the mix – especially if you can cross reference it with closed caption and subtitling of course

Gary Price, news editor of Search Engine Watch…points out that HP Labs was one of the first to produce audio keyword searches when it launched an experimental web-based tool called Speechbot in 1999. The tool cataloged more than 17,000 hours of multimedia content for everything from National Public Radio to Scuba Radio before HP took it offline earlier this month.(snip) The blinkx service is a bit more extensive. It scours more than 45,000 podcasts and already offers search for 1 million hours of TV news video and the content of academic lectures and guest speakers at the nation’s top universities.

In a few years any person publishing a podcast or doing a vlog will have to be prepared to have anyone on the planet being able to read an index that may point them directly to that bit in the file where you swear or curse the latest president or talk briefly about the latest illegal downloading tool. There will be some privacy concerns I think and no doubt various government agencies are already scouring published work for indications of subversive activity. Most good tech often comes from education or defence at some point, the commercial world just takes a little bit of it and packages it nicely for the rest of us! Still I really hope that as the text web (1.0 and 2.0 mostly) became searchable to a certain extent (see this post) then our broadband av future falls into the same findability bracket. I also like the idea of tools that everyone can use that allow anyone to enter into the personalize media game – speak forth with your voice, your opinion, press go, let everyone find you and know.

Posted by Gary Hayes ©2005