I’m sure every one of our readers is familiar with the controversy surrounding voice to text provider SpinVox. One alternative provider has come to light recently in Vox Sciences, I took a look at their service and sat down with the co-founders to find out more.
A quick look at the Vox Sciences website shows they offer much the same sort of service as SpinVox, a fully featured voicemail replacement that turns your voicemail into text and sends it to you as an SMS. Like SpinVox this is done using clever technology with a helping hand from human beings when the technology is unable to recognise the human voice.
Unlike SpinVox this is a small privately funded startup. They have less than ten employees working in a modest north London office and both the founders appear to have a solid business history. Speaking of the founders, I immediately recognised one of the names as a blast from the past, co-founder Tushar Joshi runs Lonix, the London Linux User Group which, from vague recollections of when I attended over a decade ago, was more about the beer and socialising than Linux. Tushar also runs a successful software company with clients such as Sun Microsystems while Vox Sciences co-founder Ken Blackman can reel off a long list of business successes including selling his last company to eFax.com.
So, what about the service?, first up, pricing. This is especially relevant now that SpinVox has made all UK accounts free until Christmas. Vox Sciences’ service is not free, it costs £3 per month for up to 15 transcribed messages £5 for up to 30 messages and £10 for up to 80 messages, discounts are available for bulk users or businesses requiring multiple accounts. I found this pricing scheme a little bit confusing at first but upon further investigation it seems that users are billed the full amount regardless of whether they use their allocation, for example if I pay for the £3 service and only use 5 messages in one month I am still billed the full £3. If I receive more than my allocated voicemails the billing cycle is restarted, for example if I use all 15 messages in the first five days of my subscription the system bills me another £3 for a further 15 messages (or, if I’m on the £5 or £10 tariffs it’ll repeat those amounts). While this is obviously more expensive than SpinVox’s free offering and the way the pricing scheme is presented on their website is a little confusing it works out to be a very reasonably priced service. It’s also worth noting that the new free SpinVox service doesn’t include SMS delivery for new users while Vox Sciences offer both e-mail and SMS.
Setting up the service is simple, you sign up on the website and are given a voicemail number to divert your calls to. As you’d expect you can record a custom greeting and set a PIN. Once set up, whenever someone leaves you a voicemail it gets passed through Vox Sciences’ in-house voice recognition system (called VERBS). If VERBS can recognise the speech then the text is sent to you via E-Mail and SMS otherwise the unrecognisable parts of the message are sent via a human operator who will transcribe it manually. In practise the system works well, it’s a little bit slower than SpinVox but the accuracy seems to be higher. One nice feature is that the e-mails optionally contain an MP3 attachment of the recording so you can hear the message without having to make a phone call and you have a recording that you can save if you need to. While we’re on the subject of saving audio, I asked about the company’s data retention policies and was told that the only copy of the messages that Vox Sciences keep is the one you can access via their phone or web interfaces and that once you delete the message it’s not stored anywhere else on their systems.
Perhaps the biggest issue with this type of service has always been the extent to which a human operative gets involved and the security around that. People are rightfully concerned about their personal messages and this is a very sticky area to be involved in right now. Vox Sciences are quite open about their use of third party farms of human beings (referred to by those in the industry as QC houses) and Tushar explained to me that the technology to perfectly transcribe human speech just doesn’t exist right now. Massive corporations such as Microsoft as well as government intelligence agencies the world over are trying to get it right and there are some impressive applications using limited vocabularies but there just isn’t anything possible at the moment that can adequately transcribe an entire language along with it’s regional dialects and slang as well as background noise. Vox Sciences aren’t trying to break through this barrier, they have impressive technology that can transcribe most messages left in ideal situations but when someone calls you from a busy pub or noisy roadside it falls back to humans to transcribe these messages. Tushar took the time to explain to me how messages flow through the Vox Sciences system and how they are split into small chunks of audio as they are processed. It’s this splitting of the audio that adds an extra level of security to a message, if a chunk of audio can’t be recognised it’ll be sent through to a human QC agent for transcription, meaning that the entire message is never listened to by a human. Tushar explained to me that the exception to this rule is when a UK based employee of Vox Sciences listens to your message for support purposes, for example if you have a complaint about the quality of your transcribed messages. Finally the companies entry on the data protection register reflects that data may be processed but not stored outside the EU and that all data is stored in the UK.
Both Ken and Tushar seem very excited about the future of Vox Sciences and rightfully so given that as they approach the end of their first year of trading the business is breaking even and has achieved a significant level of traction in the business community they are targeting. Thanks to their conservative approach to business you won’t find any expensive tenuous PR stunts or free social media integration. Instead you’ll hear about solid organic growth and how plans are afoot for increasing the number of security and integration options to appeal to a wider range of business users. When asked about targeting Mobile Networks as customers Ken laughed and explained that they are happy to talk to them but it’s not their goal to sell through MNOs, which makes sense to anyone who has every tried to work with an MNO in the past.
Ultimately it’s still early days for Vox Sciences but the future looks very promising and while the industry dominator is currently under fire from all quarters they have the opportunity to sneak through and steal the UK market for themselves.
Users in the UK can get a free 7 day trial of Vox Sciences from their website, VoxSci.com
Disclosure: James Whatley is a co-founder of The Really Mobile Project and also serves as SpinVox’s Head of Social Media. He had no part in the writing of this article and is currently on a sabbatical from The Really Mobile Project while he takes part in the Lucozade Energy Challenge and while the SpinVox controversy is ongoing.
Dan Lane is the author of this article and has complementary accounts from both Vox Sciences and SpinVox. If you have any concerns about bias in this article the author urges you to try both services and let us know which you prefer in the comments below.