Tuesday, July 15, 2008

You want my voicemail too?

I was recently asked by a colleague what I thought the most interesting challenge was going to be for archiving over the next couple of years. My response was voicemail archiving. With the last release of Exchange Server Microsoft has made it easier for archiving solutions to grab voicemail and store it in their archives. Since voicemail can now be part of an existing email as an attached .wav file there is no reason why an archiving solution should not be able to store this data. Storing the data is the easy part and probably not the most interesting evolution in archiving but indexing and searching that voice data is.

An Article from Wisconsin Technology News that has been sitting in my inbox for a while now recently recaptured my attention. It discussed how the evolution of VOIP technologies can raise eDiscovery concerns for companies. Since voicemail is stored as a file and unified communications solutions store this data in the email system courts are now looking at this information as discoverable.

In pure electronic form the voicemails are not very useful. Solutions on the market today that index data do so by using the text contained in files and emails. The data in an audio file would need to be converted to text first and then indexed to be easily recalled in a discovery request. The article describes three ways of grabbing these keywords for indexing:

· Phonetic – Looking at patterns in the speech to determine words (Not accurate or likely to happen in my opinion)
· Manual transcription – Sending the voice file away to be manually transcribed (Prone to human error and expensive)
· Automated transcription – Using a speech to text conversion process (Most probable solution)

All three methods for gathering text from voicemails have yet to be perfected and all still seem to have slight deficiencies. It is my opinion that it will be very expensive for companies to produce voicemail data in a discovery request until a solution that contains a suitable filter (that can be integrated into existing indexing engines) for converting speech to text (effectively and without error) can be created.

Article Link: http://wistechnology.com/articles/4789/

Technorati Tags:
, , , , , , , , , ,

2 comments:

Sam said...

Have a look at voicecloud.com. It solves the problem of archiving since everything is stored on Amazon's cloud computing infrastructure which means unlimited storage of voicemails. The messages are also transcribed so they are searchable through the online interface.
I guess you could still use Microsoft Exchange to archive voicemails since they will now come to you as an email with the .wav or .mp3 attachment and the text included in the same message.

Ron Robbins said...

Thanks for commenting...

I posted your comment so those interested can visit the website to look at the information. It looks like this would provide the automated transciption and keyword search required.

Thanks for the link...