VUI ain't GUI!

There are a lot of differences between a voice user interface and a graphical user interface - and to be quite honest VUIs are more difficult to do than GUIs because they have a lot more constraints.

VUIs are sequential - GUIs are parallel

In a VUI information can only be presented to the caller in a sequential manner. This is fine if the amount of information is small, because the caller can remember it. However, it’s a bit of a disaster when it comes to something like a train time-table. Let’s say someone wants to take a train at 8:00 to Paris. The application is clever enough to know that they may want to leave a little before 8:00 or a little after - so it offers three or four options.

Regional train 2212, leaving at 7:58, 1 change over at Rouen, arriving at Paris at 10:05. Intercity train Le Coq, leaving at 8:12, arriving at Paris at 10:10. Regional train 2217, leaving at 8:16, 1 change over at Rouen, arriving at Paris at 10:33

If you have to listen to this in one long go, it’s very hard to take in all the information. However, you could be sensible and let the caller navigate through the list, using words like next and previous. While this would doublessly improve things, they still have to be able to hold all the information in their head. Now take a look at the same information, prepared for a visual interface:

With a GUI you can peruse the information at your leisure and don’t have to burden yourself with trying to remember the whole time-table.

A VUI may be good at letting the caller input varied information, however, it’s weak point is presenting a lot of information back to the customer. A GUI is much more suited to this kind of work.

ASR VUIs are more error-prone than GUIs

It’s a sad fact of life, but ASR is rarely perfect - much and all as we may wish it were. And the more complicated the application the more likely it is that something will go wrong. That’s why UI design is so important - using error prompts and help messages in a constructive way, can help close the transaction for the caller even though there may have been ASR errors. Hence the idea of escalating error prompts, which make it ever clearer to the caller, what they may say in order to be recognised.

VUIs offer poorer feedback than GUIs

When things go pear-shaped with loading a webpage you can try reloading it. You can go back and then try opening the page again. If you think the browser is dead, you can restart the browser and try again. As a matter of fact, you probably aren't even surprised when a page doesn't load first time. Servers being under load and not being able to cope with demand is nothing new in the internet world. If push comes to shove, you can always try again later.

Unfortunately that's not the way things can be done in the telephony world. There's no status bar telling the caller how much of the page has loaded. You can't even create an acoutic status bar, because in the application you don't actually know how far along the process is. So time is a critical issue in VUIs - processes have to work fast enough that the caller doesn't have the impression that the service has hung and they're stuck listening to music with no prospect of getting any further. Similarly, as there is no reload button that the caller can press, the application can in theory do a reload - while asking the caller to wait a little longer. However, it would be better if the process were to work first time everytime.

If you have any comments, ideas, issues, etc. about this topic why not try the voice-push forums