VoiceCon: Avaya Converts Voicemails To Text With SpinVox

Avaya has partnered with communications app development company Mutare Software and speech technology company SpinVox to translate voicemails into text, the company announced this week at VoiceCon. The software, Avaya Voice to Text, works with Avaya's Modular Messaging unified messaging system to push transcribed voicemails out to employees as e-mails with optional audio file attachments.

Avaya automatically submits voicemails to SpinVox's Voice Message Conversion System, which converts speech into text via machine, rather than relying on human transcription. Once the message is transcribed -- the process typically takes two to five minutes -- SpinVox sends the message to the employee's e-mail address along with an optional audio file. The service can convert voicemail to text message and multiple document types as well as e-mail.

Since they're now in text form, messages can be easily saved and searched, turning old voicemail messages into a potentially valuable piece of content. Names, numbers and details that might have required playback of a message to remember or that are difficult to catch and write down the first time around now get automatically transcribed. And since the messages are both in e-mails and part of Avaya's unified communications software, employees can click to either respond via e-mail or call the sender back. Employees control how they want to receive their messages: as a daily batch, or as they come in.

But converted voice mails could be used for more than just employee reference because text is much easier than audio to integrate with other business processes. "A unified communications platform can really enhance SpinVox and provide deeper functionality," Todd Woodstra, SpinVox's Vp of unified communications and emerging markets, said in an interview.

For example, phone survey results could be automatically compiled, saved and analyzed. Voicemail audio and text could prove a godsend during e-discovery or in the event of a compliance audit where record retention rules apply. Calls coming into contact centers could be routed or escalated based on keyword searches, similar to how today's IVR systems work, but in this case, calls coming in after closing hours could be automatically routed and returned by the appropriate agent in the morning.

Though Woodstra is confident of SpinVox's accuracy -- once-difficult numbers, money and dates are improving rapidly in accuracy, he said -- speech-to-text remains an imperfect art. SpinVox continues to build a library that includes things like grammar rules, proper names and anticipated caller behavior patterns.The trickiest calls, where SpinVox's transcriptions don't meet certain automatic "confidence scores" that measure a transcription's expected accuracy, get additional transcription support from a team of people, rather than simply relying on machine transcription. "There are still certain things that a speech recognition system wouldn't understand," Woodstra said.

Until now, SpinVox, founded in 2003, has primarily focused on the carrier market. It has partnerships with Alltel (which costs consumers the low price of $5.99 per month for unlimited transcription), Cincinnati Bell, Rogers, Vodafone Spain and a number of other mobile carriers, as well as Skype and LiveJournal. Deeper integration with Facebook and Twitter is coming soon.

However, it has also done some limited engagements with businesses outside of the Avaya partnership. For example, a large financial company has committed to subscribing to the service. Though a company using SpinVox today outside of the Avaya integration will likely have to do some custom coding to make SpinVox work with its voicemail system, SpinVox is actively seeking partnerships with other telecom and unified communications vendors.

Traditionally, SpinVox's service has been sold on a per-conversion model that comes out to much more than Alltel's price, but that could change. SpinVox is one of several speech-to-text vendors for voicemail, along with CallWave, PhoneTag and GotVoice, and the technology continues to improve, as more general speech-to-text and speech recognition features are showing up in such places as Windows Vista and Dominos pizza delivery calls.