The Hindu Business Line : The Web's a call away

IT is Monday morning and you are driving to the airport to drop your daughter. The radio announces that due to bad weather there is a traffic jam. While you drive, you use your cell-phone to `call' the airline's Web site and do a telecheck.

Does this scenario appear improbable? Not at all! It is possible to access the Web by dialling, instead of typing with a keyboard, and talking, rather than clicking with a mouse.

`Voice Portals', also called `Speech Portals', allow one to access the Web via the telephone. They place the wealth of the Web at `lip tip', facilitate physically-challenged users and those with limited motor skills to use the Internet, and provide enhanced customer service via Web-centric, self-service applications.

This self-service can include general information retrieval - such as stock quotes, flight information, weather, directory assistance; unified messaging - such as `listening/talking' your e-mail and receiving pager messages via telephone, etc.

Interactive voice response or IVR applications used to provide customer service have limitations of sequential access and often confusing long menus. Most importantly, they provide limited information only.

How do voice portals work?

A typical Web page is written using HTML. In voice portals, VXML (or SALT by Microsoft aficionados) is used. The text embedded between VXML tags is converted into voice. The phone number from a voice gateway is mapped to the URL where the VXML page resides. Thus when a user types a URL or clicks a link, the request is sent to the Web server hosting the site and the page is loaded in the user's browser.

In voice portals, the user dials a phone number and the request is sent to a media server that presents the information in the form of voice. A gateway is connected to the Internet and the public switched telephone network. This can process both voice and IP packets.

VXML applications use a combination of speech and touch-tone commands to access different parts of the portal. The output can be synthesised speech or digital pre-recorded audio. Using VXML is akin to filling an HTML form. The user has to choose among the options provided similar to choosing from a dropdown box in HTML form. The other component that is unlike HTML is that in order for the VoiceXML interpreter to understand what the caller is saying, grammar text file is needed.

Speech Grammars help programmers frame rules covering the various combinations of replies that users may provide to a particular question/context. For example, a user may reply saying, `yesterday' or `5th' or `Sunday', all referring to the same.

What is the difference between voice dictation and VXML? In voice dictation, the software has to be `trained, tuned and tested' to recognise and process the users' voices. But, telephone voice recognition must work without training, on poor lines and noisy cell-phones, and for many callers with differing dialects and accents, all at the same time. Therefore, voice recognition is based on grammars that increase recognition accuracy by limiting the things people can say. Grammars are defined using a grammar language; either the proprietary GSL Nuance format, or the w3c SRGS (Speech Recognition Grammar Specification) format.

In voice portals, navigation is done by speaking, routers and the Web logs by phone number records. However, there are a few limitations of voice portals. Images cannot be viewed, one can only listen to their descriptions. Animations with impressive colours and displays cannot be seen.

The telephone call can be sent via plain old telephone system or via the Internet (VoIP - Voice Over IP). IP-managed networks allow voice, video and data to be transmitted via a single wire with guaranteed quality of service, at higher speeds, volumes and reliability.

The current trend is to use IP for all new networks. For VoIP applications, Session Initiation Protocol (SIP) is used. These applications are emerging fast and rapidly penetrating into the business world.

The author is Consultant, Satyam Computer Services.

Article E-Mail :: Comment :: Syndication :: Printer Friendly Page

Stories in this Section
Life cycle of a captive BPO

The buzz is louder

Doing IT differently

Sootaspeak

The Web's a call away

RAM-induced problem

Removing virus program

Drive allocation

Ready for the opportunity

Fencing out trouble

Monitoring Net traffic

Quiz

Create work and then sell it to someone else

Cartoon