Ipadio Selects SpinVox to Power Android App

Ipadio has chosen voice-to-text provider SpinVox to power the speech recognition capabilities on its new mobile application for the Android platform.

Ipadio is an audio-to-Web service provider that allows users to record and post audio content onto blogs and social media services from any phone. The company also offers a commercial version with workflow features, like moderation, group access and distribution, and save to database, as well as additional in-call functionality. With Ipadio’s use of the SpinVox application programming interface (API), however, the audio-based services will be augmenting all of those functionalities on Android with discoverable and searchable text. 

The SpinVox engine will be used to automatically generate titles, tags, and a transcripts to sit alongside the audio. The results can be integrated into social media accounts such as Twitter, Facebook, Wordpress, and Blogger.

The SpinVox partnership comes after an extended search process on Ipadio’s part.

“Voice delivered via telephony or smartphone can be contextualized,” says Mark Smith, Ipadio’s CEO. “You could put it in embed or in part of a blog, but it’s not easily indexable. Search engines are not yet sophisticated to pluck out the words from voice.”

What Ipadio needed was text that could be more easily searched by traditional engines. When Smith began his search he was initially disappointed. He found that many vendors had speech recognition platforms, but those systems required training. The ones that didn’t, moreover, returned “reasonably disappointing” results. Unable to provide a reliable full transcript, Ipadio began looking into using an engine to associate recordings with “two or three top terms”—tags. In April, however, SpinVox launched its API, which later came to Smith’s attention.

“We ran some tests and, I have to say, I was surprised by the results,” Smith says. “The results were so good that we…could not only supply tags, but full transcripts. With the degree of accuracy that SpinVox provides, it’s possible to add considerable business intelligence around audio submissions.”

Praise of SpinVox’s engine likely comes as welcome news at its U.K. headquarters. In recent days, the company’s service has come under criticism and speculation in the British press since the BBC published a news story and a blog post last week suggesting the voice-to-text provider might use live people to do the bulk of its transcription, rather than just serving in a supporting role as the company has claimed. The reporting relied heavily on charges made by employees of a call center in Egypt contracted to do work for SpinVox. 

SpinVox and its supporters, for their part, have roundly and swiftly denied the BBC allegations, citing a number faults with the BBC’s sources—chiefly that the Egyptian call center never made it past SpinVox’s vetting process, never actually worked on live messages, and that the individuals who spoke with the BBC would not have been in a position to accurately comment about the company’s process. SpinVox also maintains that a majority of its messages are transcribed automatically, with no human intervention. 

Smith himself is eager to join in supporting SpinVox’s recognition engine without prompting. “Our engagement with SpinVox has been extraordinarily positive; they’re a deeply ethical company,” he claims. “I’m somewhat disturbed by compatriots’ willingness to give them a pasting for something I don’t understand.”

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues