VoiceXML Applications

Here's a very small list of some of the applications that you can create with VoiceXML.

Replace existing IVR Services

All of the more basic DTMF-driven IVR services can be re-created with VoiceXML and the open service architecure that it espouses. The caller isn't going to notice much of a difference unless you take the migration as an opportunity to rework the user interface. However, moving from a proprietory IVR to a more flexible modern architecure does bring a lot of technical advantages. The advantage that VoiceXML offers is the speed with which changes can be made and new functionality added. You are no longer dependent on your IVR supplier to make any modifications you require and you can re-use existing IT infrastructure and interfaces. These engineering changes have an effect in terms of the product, as any modifications are now easier and quicker to do, resulting in a faster time2market for better customer-oriented services.

Actually, by randomising some prompts and re-thinking the caller experience, it is often possible to rejuvenate tired IVR applications. Having said that, you need to be a little careful. If there are a lot of power users, who just press a series of DTMF keys, without listening to the prompts, in order to get to a particular point in the IVR tree, then you may need to keep the basic option-layout the same. Another possibility would be to play a prompt at the start which has no barge-in, which informs the caller that the IVR application has changed. Typical examples of such IVR applications are pre-paid mobile top-up, basic call-centre call routing, etc. In general these applications are more common in North America than they are in Europe (as are speech recognition applications).

Voice Portals

Anybody remember the great rush of voice portals about 5 or 6 years ago? TellMe, Hey Anita!, BeVocal, AudioPoint (now defunct), Eckoh were all showing off voice portals. Listen to the weather, the news, the sports results, the traffic, your emails, your calendar, ring people with your contact address book.

Oh, what a bright future!

Unfortunately, nobody really needs a voice portal. It's not that they don't work - they do. But we are continually bombarded by information all day, whether it's email, SMS, web, television, or telephone calls. Sure there are occasions, where you've no internet connectivity, or your Blackberry isn't working. But they are so few and far between, that they don't make a convincing business case. Having said that, developing a voice portal is a great way to learn VoiceXML and speech recognition - just don't expect to make millions with it!

Name Diallers

Name diallers are useful - and relatively simple - you say the name and you get connected. Pretty simple. The application can be used internally in a company, as an after-hours receptionist, for all external calls, and with the right kind of tuning for directory assistance inquiries. The latter obviously are in a league of their own as regards the number of names to be recognised - typical enterprise diallers have to handle anything between fifty and several thousand names - but nothing in the millions as in the case of directory assistance. Actually name diallers are ideally suited to small & medium sized enterprises (SMEs), as it is cheaper than a receptionist and the recognition rate is very good for a smaller number of names.

The catch with name diallers is that you need a really good dictionary of names. In particular foreign names cause the most trouble. The reason for this is that the recogniser, if it can't find the name in the dictionary, uses rules to determine how the name should be said. While this works well for names in the same language, it rarely works well for foreign names. So if you don't have an amazing dictionary covering all possible names, then you might need to consider, offering the customer the chance to enter some names phonetically themselves. They don't have to be able to use the SAMPA alphabet to do this. Quite often it's enough to write it the way it sounds - i.e. a "sounds like" version. It's even better if you can offer a TTS engine that uses similar rules as the recogniser for pronouncing words. Then you can listen to the "sounds-like" version of the name as the TTS engine interprets it and tell if it's ok or not.

Call Centres

This is one area where the business case for a speech recognition application is easily justified. Call centres suffer from a high turn-over of staff, not least due to the mind-numbing aspects of much of the work. Speech rec apps can be used to take over some of the more mundane work, leaving the call centre agents to concentrate on more challenging tasks. So although the cost of writing a speech rec app is quite high, the minutes saved in the call centre quickly pay it back. These applications are becoming more and more common - not least due to flexibility of implementing them with VoiceXML.

Outbound Calling

This is a rather untapped market. You want to make your customers a special offer? You need to remind them of something (they have an appointment tomorrow, their credit card bill needs to be paid, their pre-paid phone credit is about to run out)? Then why not call them? You can use VoiceXML for automated outgoing calls as well. The only thing to watch is your algorithm to decide when to try calling someone again - you don't want to end up annoying your customers by hassling them with calls all day. Other than that, outgoing calls seem be as good as, if not better than text messages for marketing campaigns.

VoiceMail

Yes, even humble old voice mail can be VoiceXMLified. Using the VoiceXML layered architecture allows voice mail to become an integral part of the IT landscape, similar to email. As a matter of fact, modern voice mail systems are using email stores to save the messages in. Video mail takes a similar approach. With this approach to voice mail, the once ubiquitous phrase Unified Messaging is finally beginning to make some sense.

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