Brad's Books and Organizations

Books

Books

Organizations

Organizations

Hi, I’m Brad Feld, a managing director at the Foundry Group who lives in Boulder, Colorado. I invest in software and Internet companies around the US, run marathons and read a lot.

« swipe left for tags/categories

swipe right to go back »

Outsource Your Manual Stuff

Comments (10)

I didn’t listen to any voicemail in 2007.  I got plenty of them – but I didn’t listen to any of them.  Yet I got 100% of the information that people left to me.  I discovered a magical system that transcribes voicemails to emails.  And it doesn’t involve anyone that works for me.

I’ve been searching for a speech-to-text voicemail conversion system for a while.  I approached the problem incorrectly – I assumed that it was a technology issue.  I’d periodically come across something that tried to do what I wanted but truly sucked.  For a while I had my assistant transcribe my voicemails, but that was tedious, a poor use of her time, and often had a lag if the call was after hours or when she was in the middle of other stuff.  Plus, she hated doing it (think "extremely tedious part of the job – yuck.")

In my quest for a solution I came across SimulScribe early in 2007.  It’s exactly what I want.  For $30 / month I have my voicemails immediately transcribed and sent to my email account with the text and an attachment of the message (if – for some reason – I have to listen to it to pick up a nuance that the transcription missed.)

The real aha was that SimulScribe didn’t actually try to solve the hard technology problem (in this case – translating voice to text.)  They simply outsourced it to India.  Transcription has been around for a long time (it’s what your doctor’s office has been doing for decades) – using the Internet, email, low cost outsourcing, and a simple web service makes this a marvelous solution for the voicemail haters of the world.  I wrote about an analogous dynamic on the post Manually Automate Your APIWhile it’s nice to believe that technology can be applied to processing any and all data, there is an old fashioned way to get there quickly.  In some cases, like SimulScribe, it’s quite elegant.

Other folks are using this same approach, including SpinVox and Jott.  I’m been happiest with SimulScribe for voicemail and can’t imagine living without it.  I look forward to 2008 being another voicemail free year.

  • http://www.spinvox.com Jonathan Simnett

    The above piece gives the impression that Simulscribe achieves its voice to text service through outsourced human transcription in India. It also goes on to say SpinVox – thanks for the mention – and Jott are `using the same approach`

    This is not the case with SpinVox. SpinVox has developed a sophisticated learning system called the Voice Message Conversion System or D2 which carries out automatically by computer the majority of SpinVox voice to text conversions. Messages are converted by machine; humans are only involved to help train the intelligent system to improve its speed and accuracy (although SpinVox users do report already 97 per cent accuracy where other automated services deliver only the gist of what is said).

    So, when the system encounters a word or phrase its does not know or understand, it is able to refer to a specialist in the SpinVox language laboratory for assistance. The human then trains the system so that word or phrase becomes known to the VMCS for future use. In that way, the VMCS is constantly evolving and learning, increasing in accuracy and speed with each conversion.

    • http://www.feld.com Brad Feld

      Jonathan – thanks for the correction. I just took a look at the website again – do you have an unlimited plan?

    • Tarun

      Jonathan – Interesting comment. All you are saying is that automated systems are used and manual transcription when necessary. Even SpinVox has BPOs all over the world including Windies, South Africa and India!! Trying to masquerad them as a laboratory and fancy words doesnt cut it :-) Well, net net you are using the same approach as others though your recognition technology could be better.

    • Jonathan Simnett

      Brad – if you subscribe to the SpinVox service direct from the SpinVox site then there is currently no unlimited plan, however some carriers offering the SpinVox service have launched unlimited plans. Customers of Rogers Wireless, for instance, can have the Rogers Voicemail to Text powered by SpinVox service added to their plans in available markets for $15/month for an unlimited number of voicemail to text conversions. The first month of service is free. Cincinnati Bell offers unlimited conversions for $19.99 per month; Sasktel also offers an unlimited plan for $25 a month.

  • Jim Pollock

    I ran across these guys last year when David Cohen tried it out. Very interesting how a hosted (centralized resource like voicemail) solution leverages normally linear human resources into a scalable solution. I looked at the same model for near-realtime language translation capability for a bizplan a couple of buddies and I were contemplating a couple of years ago. Jim

  • Saul Lieberman

    I was curious to see how SimulScribe’s website explains its service.
    Maybe I missed it, but … I found no reference to human intervention, in India or otherwise. On the other hand, their home page says “SimulScribe utilizes cutting edge voice recognition technology to convert your voicemail messages into text.” That might be a true statement but it omits a material fact necessary to make the statement not misleading.
    I would imagine some of SimulScribe’s customers would not be happy to hear that someone in India is listening. (And some, perhaps most, customers do not mind).

    • http://www.feld.com Brad Feld

      I personally don't have any trouble having it transcribed by someone else but it's interesting that the web site doesn't represent it this way. I wonder if I'm wrong and they have a core algorithm for this (although I can't believe that's the case given the results.)

  • Dave

    Almost forgotten in this post is the fact that you get all your voicemails in email, therefore the people who think they're somehow getting your attention more effectively by calling are wrong… presumably, the only reason to leave Brad a voicemail is because one is driving and can't type, or some such.

    It would be interesting also to hear your impressions of seeing voicemails in email. I bet a large fraction of the content is spurious. However, I can picture it being useful for someone to leave you a longer, substantive voicemail in this scenario because you just read it instead of having to sit there and listen to it.

    • http://www.feld.com Brad Feld

      Yes – much of the content is spurious – I much prefer emails. However, the window of time for the voicemail is limited (I think to 60 seconds) so longer ones (thankfully) get cut off. And – I can read a 60 second voicemail in about 10 seconds!

  • E Sack

    I actaully like http://www.gotvoice.com/ better than Simulscribe you might want to try that one out too.

  • Jonathan Simnett

    The above piece gives the impression that Simulscribe achieves its voice to text service through outsourced human transcription in India. It also goes on to say SpinVox – thanks for the mention – and Jott are `using the same approach`

    This is not the case with SpinVox. SpinVox has developed a sophisticated learning system called the Voice Message Conversion System or D2 which carries out automatically by computer the majority of SpinVox voice to text conversions. Messages are converted by machine; humans are only involved to help train the intelligent system to improve its speed and accuracy (although SpinVox users do report already 97 per cent accuracy where other automated services deliver only the gist of what is said).

    So, when the system encounters a word or phrase its does not know or understand, it is able to refer to a specialist in the SpinVox language laboratory for assistance. The human then trains the system so that word or phrase becomes known to the VMCS for future use. In that way, the VMCS is constantly evolving and learning, increasing in accuracy and speed with each conversion.

  • http://intensedebate.com/people/bfeld bfeld

    Jonathan – thanks for the correction. I just took a look at the website again – do you have an unlimited plan?

  • http://intensedebate.com/people/saul_lieber4040 saul_lieber4040

    I was curious to see how SimulScribe’s website explains its service.
    Maybe I missed it, but … I found no reference to human intervention, in India or otherwise. On the other hand, their home page says “SimulScribe utilizes cutting edge voice recognition technology to convert your voicemail messages into text.” That might be a true statement but it omits a material fact necessary to make the statement not misleading.
    I would imagine some of SimulScribe’s customers would not be happy to hear that someone in India is listening. (And some, perhaps most, customers do not mind).

  • Dave

    Almost forgotten in this post is the fact that you get all your voicemails in email, therefore the people who think they're somehow getting your attention more effectively by calling are wrong… presumably, the only reason to leave Brad a voicemail is because one is driving and can't type, or some such.

    It would be interesting also to hear your impressions of seeing voicemails in email. I bet a large fraction of the content is spurious. However, I can picture it being useful for someone to leave you a longer, substantive voicemail in this scenario because you just read it instead of having to sit there and listen to it.

  • http://intensedebate.com/people/bfeld bfeld

    I personally don't have any trouble having it transcribed by someone else but it's interesting that the web site doesn't represent it this way. I wonder if I'm wrong and they have a core algorithm for this (although I can't believe that's the case given the results.)

  • http://intensedebate.com/people/jim_pollock555 jim_pollock555

    I ran across these guys last year when David Cohen tried it out. Very interesting how a hosted (centralized resource like voicemail) solution leverages normally linear human resources into a scalable solution. I looked at the same model for near-realtime language translation capability for a bizplan a couple of buddies and I were contemplating a couple of years ago. Jim

  • http://intensedebate.com/people/bfeld bfeld

    Yes – much of the content is spurious – I much prefer emails. However, the window of time for the voicemail is limited (I think to 60 seconds) so longer ones (thankfully) get cut off. And – I can read a 60 second voicemail in about 10 seconds!

  • http://intensedebate.com/people/tarun4062 tarun4062

    Jonathan – Interesting comment. All you are saying is that automated systems are used and manual transcription when necessary. Even SpinVox has BPOs all over the world including Windies, South Africa and India!! Trying to masquerad them as a laboratory and fancy words doesnt cut it :-) Well, net net you are using the same approach as others though your recognition technology could be better.

  • E Sack

    I actaully like http://www.gotvoice.com/ better than Simulscribe you might want to try that one out too.

  • http://intensedebate.com/people/jonathan_si4080 jonathan_si4080

    Brad – if you subscribe to the SpinVox service direct from the SpinVox site then there is currently no unlimited plan, however some carriers offering the SpinVox service have launched unlimited plans. Customers of Rogers Wireless, for instance, can have the Rogers Voicemail to Text powered by SpinVox service added to their plans in available markets for $15/month for an unlimited number of voicemail to text conversions. The first month of service is free. Cincinnati Bell offers unlimited conversions for $19.99 per month; Sasktel also offers an unlimited plan for $25 a month.

Build something great with me