Talking To The Xbox
This Friday will see the launch of the Xbox One, which features built-in Bing to power its searches –and a whole new way of using it. This will be the first Microsoft product to let users search simply by speaking. It is powered by deep neural network technology, helping it to understand natural questions – but can it deliver?
“When we watched Star Trek decades ago, the idea of talking to a computer seemed like pure science fiction,” says the Bing blog. This may come as surprise to those who have been shouting at their computers for decades, but the machines were not generally expected to react. Bing goes on to say: “Fast forward to the present and your voice has replaced the remote control for your living room with Xbox One, an all-in-one games and entertainment system with sophisticated voice navigation and natural language voice search powered by Bing and Kinect.”
The Xbox One is primarily geared toward entertainment and this is likely to limit the kind of questions it is usually asked, making the e questions easier to anticipate. It may be worth thinking about for those promoting entertainment-related products, as previous voice-based systems have run into difficulties with the kind of acronyms and deliberately unusual spellings that are common in the entertainment industry. At this stage it is not clear whether the system will become familiar with its user’s voice or rely on a general set of expected phonemes; if the latter, it could run into similar problems to those that Siri has run into when it encounters unusual accents, in turn creating major problems for some users.
At the core of the Xbox One’s search system are specific command phrases that must be used to introduce each question. There are a number of these and they seem flexible enough to cover most eventualities. More problematic, though, is that the Xbox One is always listening, even when in standby mode, and its command phrases are equally likely to come up in natural speech. This could lead to a lot of automatically triggered searches for things such as a cup of coffee or more sleep.
Bing says that it has spent years working to resolve common problems in other voice-based systems. Whether this will prove useful in the real world remains to be seen; however, if it does, it is another step on the road to more fully-integrated systems.
