Mike Butcher at TechCrunch is reporting that Amazon appears to have bought mobile virtual assistant developer EVI (formerly TrueKnowledge) for $26 million. Butcher was unable to confirm the story through either Amazon or Evi spokespeople, but, as we discovered in our long-standing coverage of Amazon’s acquisitions in the speech processing domain, stony silence is the order of the day. In November 2011, the Web’s largest retailer and cloud services provider very quietly acquired Yap, which had been offering highly-accurate speech recognition and transcription. In January of this year (2013) it expanded its automated speech processing chops by purchasing text-to-speech specialist Ivona, which had built its reputation by providing highly-human sounding renderings of long-form text.
While TechCrunch’s Butcher speculates that the next step for Amazon, given its serial acquisitions, would be a telephone. My own belief is that it is poised to make all its browser-based, e-commerce assets more conversational. I believe the most common use case would be a more natural interface for the Kindle, whether it’s thought of as a mobile device or a mobile application running on iOS or Android. The idea would be to provide a consistent, highly personalized user interface across all potential access points. Thus a “talking Kindle” (or Conversational Kindle) would be the entry point for everything that Amazon has to offer. That includes video entertainment, books and the wide variety of goods and services offered through Amazon Prime, the Kindle Store or any of the online merchants supported by Amazon.com. Offering a Kindle phone makes less sense to me, but the option of supporting person-to-person communications (phone conversations, conference calls or video conferencing) is not out of the question.
UK-based EVI was conceived as a company called TrueKnowledge in 2007 and built its original intellectual property and productization efforts were based in becoming “The Internet Answer Engine.” This involved building a knowledge base of structured “facts” which could be used to respond to user queries. It later launched EVI as a speech-enabled counterpart to Siri. When Opus Research tested the service a couple of years ago, we found recognition of our spoken words were often misunderstood, but performance issues of this nature often improve over time, as the company builds a better corpus of user utterances.
Integration with Amazon could give EVI a performance boost as Yap’s recognition engine and Ivona’s text-to-speech rendering are brought online. Positioning EVI as a speech enabled personal assistant to Amazon.com will enable designer, developers to have a better sense of the context of a query. That is to say, knowing that a visitor is browsing for books automatically limits the range of responses that Amazon needs to provide. Ditto for searching and selecting filmed entertainment or other product categories.
When it comes to cloud-based commerce and communications, Amazon is in the category with Google, Apple and Microsoft. So far, only Apple has made the bold move into develop and market a speech-enabled mobile virtual assistant, the archetypal Siri. As I learned at Bill Meisel’s Mobile Voice Conference earlier this week, Google puts more emphasis on making “voice” an option for everything that it does, without developing a “virtual assistant” a la Apple’s Siri or Angel’s Lexee. Microsoft’s emphasis, since it spun off Tellme to [24]7 is on a “natural interface” for Web-based resources including spoken words and gestures – with less emphasis on human-like interactions. Among them, Amazon is a leader in creating a highly personalized shopping experience, with knowledge of a buyer’s past history, one of the longest-standing “recommendation engines,” and even formal support of each member’s “Wish List.” Imagine Apple’s Siri having the benefit of an individual volunteering information about what they are looking for in advance.
Making a smartphone, to compete with Apple’s iPhone or the broad variety of Android devices, would not play to Amazon’s strengths. Building a speech-enabled, virtual personal assistant is highly-personalized and responsive to each user’s intent is a different matter altogether.
Categories: Articles