[brog] The Future of Audio Search is Here!

Posted by on February 17, 2006 at 9:56 am.


One of my portfolio companies, CallMiner, just announced a grid-based indexing platform that will forever change how audio is mined in the enterprise, government and internet. The historical battle between processing power and audio indexing has been decoupled by CallMiner’s Virtual Server Room (VSR) platform, allowing for higher accuracy and virtually eliminating hardware constraints for speech analytics and audio search. [Not to mention finally harnessing grid with a business model, sorry SETI]

What does this mean for speech analytics?
It means that enterprises who weren’t mining 100% of their contact center recordings because of the hardware burden can now unleash the business intelligence hidden in every one of their customer calls. When customers are saying “competitor X does this better”, “I would pay more for your product/service if it did Y” or even “I’m going to switch to competitor Z because your customer service sucks”; enterprises can now know and react to those comments “on the fly”. With all of the effort/dollars plowed into CRM/BI systems, shouldn’t the single largest customer datapipe in an enterprise (their call center recordings) be tapped to deliver immediate value? Absolutely and VSR makes that a reality for all…

What does this mean for audio search?
It means that other datapipes of audio can also be unleashed — think audio.google.com for podcasts, vlogs, other internet audio files and possibly even VoIP. Today’s internet search engine approaches for audio come in three varieties:

1) search engine X says “it’s too hard, I’ll forego that market until it’s too late to catch up”,
2) they search audio/video transcripts (e.g. commercial broadcasts) and thus miss the mountain of audio being created as part of Web 2.0 or
3) they try to search true audio but fail on accuracy, speed or reach, and watching their “files mined” ticker is like watching grass grow as they index slower than new files are being created.

The only long-term solution is mining audio on a massive scale, but there are thousands if not millions of very large audio files hitting the net weekly. Today’s server/hardware-intensive solutions are broken. The battle between accuracy and processing power must be solved — hello VSR.

What does this mean for CallMiner’s competitors?
They can become partners. VSR is a processing platform, architected for audio — with the potential to decouple hardware constraints for other audio search/mining engines. Entrepreneurs/VCs often look to change the world and VSR does that for CallMiner and others in the search/analytics space…

If I sound pumped, it’s because I am. The CallMiner team’s proprietary innovation heralds a new age of speech/audio mining and I’m just glad to be along for the ride…

Thanks for indulging this brog (brag by blog) — I save it for only the really good stuff…

Leave a Reply