The latest downloadable version of Google’s Chrome browser showcase how Google Translate can support spoken input and text-to-speech rendering of translated phrases. As described in this post on the Google Chrome blog, the new browser supports “speech input through HTML.” In other words, a small microphone icon appears in the lower right corner of the text input box. Pressing the icon causes a “Speak Now” button to appear and enables spoken input through a PC or other device’s microphone. Captured utterances are translated in Google’s cloud and rendered as text on the Google Translate Page. Users can also choose to have the page read aloud through the browser.

Google will continue to showcase a wide variety of cloud-based speech processing applications, including speech recognition, machine translation and text-to-speech rendering. It has a huge (and growing) corpus of utterances in a multiplicity of languages which serve as the bases for ever-improving performance and accuracy. It is a source of competitive advantage to Google (albeit for services that have no direct revenue-generating capabilities for the search giant). The constant stream of innovation creates more opportunities for the general public to have a positive experience with speech recognition, as noted by Farhad Manjoo in this recent article from Slate.com, “Google has developed speech-recognition technology that actually works.”

That said, I don’t think that Google has laid to rest the ongoing debate regarding whether speech processing, translation and rendering should take place “in the cloud” or “on the device.” The browser/cloud combination makes for a good, ubiquitous demonstration platform for good quality speech processing that can incorporate large lexicons and dynamic data. But it does not solve the major issue of what to do when the device is not able to connect to the Internet. Info and resources in the cloud are of no use in that instance no matter how refined the browser is.

Categories: Articles

Tags: speech processing, Google, HTML5, translation