Watch out, Siri! AT&T on Thursday announced its intentions to compete in the burgeoning market of speech recognition technology by opening up its own, long-in-development system, AT&T Watson, to third party developers.
“In June, we plan to launch several AT&T Watson SM Speech application programming interfaces (APIs) that developers can access to quickly create great new apps and services with voice recognition and transcription capabilities,” wrote John Donovan, an AT&T senior VP of technology and network operations, in a blog post.
As he elaborated:
“The first AT&T WatsonSM Speech APIs will be focused on seven areas: web search, local business search, question and answer, voice mail to text, SMS, U-verse electronic programming guide, and a dictation API for general use of speech recognition.”
Watson, which was developed by AT&T Labs, the telecom giant’s research arm, is already live in its automated customer service phone banks and AT&T Translator and Yellow Pages mobile (YPMobile) iPhone apps.
The new AT&T Watson third-party development platform will support apps development on Android and iOS, and supports transcription, word identification and “takes spoken or written language and translates it into another language you select,” according to Donovan.
The company’s goal: create its own app ecosystem filled with all sorts of apps which harness speech recognition for different purposes, from games to lifestyle to planning to automotive interfaces — anything that creative apps developers can cook up. AT&T hopes it will have a whole store full of online apps that will overlap with the popular Apple App Store and Google Play. Check out the ambitions in the following video.
Following AT&T’s announcement, some key questions remained, such as: How much the program will cost developers, whether AT&T will take a cut of sales, and just which specific languages the technology will support.
Regarding the cost, “there is a registration charge of $99, which will allow developers to use all AT&T APIs, including speech, without a per transaction charge through 2012,” AT&T’s spokesperson told TPM. “There is a limit of 1 transaction per second. If there is an application that estimates needing more capacity, we just ask that they reach out to us to discuss their needs and we will be happy to talk about options to support them.”
As for how many words Watson can identify, AT&T pointed out that there will actually be seven different application programming interfaces, or APIs, each which contains its own “vocabulary.”
“The API belonging to the largest speech package is the Dictation (generic) API, which has been trained on 1+billion words, and supports nearly 2 million different vocabulary words,” AT&T’s spokesperson said, noting that it “can automatically detect and transcribe English and Spanish language. We have 8 other languages in the lab that we will roll out over time.”
As for how it competes with Apple’s own iPhone Siri voice recognition technology, AT&T didn’t pull any punches, with the spokesperson telling TPM:
“AT&T Watson has been used in commercial applications for over two decades. AT&T was the first to employ the largest nationwide operator assisted service in the early 90s. AT&T Watson has also been adopted by many partners including Vlingo who have publicly claimed that AT&T Watson is faster and better than other available engines.”
You’d be forgiven for thinking those are fighting words. Still, AT&T wants to play nice with Apple, at least nice enough to remain included in the App Store.