Yesterday Apple confirmed the rumor that they are integrating an enhanced speech recognition system on the iPhone. The iPhone 4S will be equipped with “Siri”, which is being described as an “intelligent assistant” to help you get things done such as check the local weather forecast or find the nearest Chinese restaurant.
Siri is being touted as a “game-changing” advance in user experience. Jason Schwarz on Seeking Alpha calls it “the most significant upgrade to the Apple ecosystem since the initial launch of the product back in 2007.” But will Siri actually bring about the revolution in speech recognition that the industry has been seeking all these years?
The implications for automotive HMI are clear: if Siri works then this can provide a very natural way for consumers to accomplish tasks such as local search and text messaging while driving. Our own research has shown that too many consumers are trying to do these tasks on their mobile phones while driving, which of course is highly distracting and dangerous due to the visual-manual distraction associated with text input. Being able to accomplish the input portion with one simple voice command can alleviate the distraction.
All of this is done using natural language processing, which has been talked about for years but hasn’t yet fulfilled its potential due to accuracy issues. Can Siri really make it “just work”? If so, we may be at the cusp of a revolution in speech recognition. We’ve found through our research that consumers are very interested in using voice, but for the most part have been extremely disappointed in the available solutions. All they need is one that “just works”, and I predict consumer acceptance and usage would take off.
Siri already has a few things going in its favor. The main issue consumers have with speech recognition systems is that they don’t know what the correct commands are. All too often, their initial commands do not coincide with available grammar, and they quickly become lost. Since Siri does not have fixed grammar or a fixed top level menu, users can vary widely in what they say. While this issue is avoided, Siri still has other obstacles to overcome:
- First and foremost, Siri has to be extremely accurate. Consumers will tolerate some errors, but frustration will quickly set in, and users will abandon the technology after just a few failed attempts.
- When errors do occur, users need to know how to exit. If an utterance is misrecognized, users will tend to repeat their command, no matter where the system took them, which typically only results in more errors.
- Users will have to adapt to natural language. Given the increase in IVR systems more consumers are familiar with speech recognition, but it is of the command-and-control variety. We have seen users hesitant to talk to speech recognition systems naturally and have felt more comfortable with providing concise commands and being walked through menus step-by-step.
Apple has already shown they can revolutionize the user experience in mobile devices. If Siri works, the automotive industry will need to partner with Apple to bring this into the vehicle. Strategy Analytics Automotive Consumer Insights service will publish a report evaluating Siri in the automotive environment after its release.