Speech recognition is a technology that has been around for many years. It was brought sharply into the public view in the late 1960s with the release of the film of Arthur C. Clarke's "2001 – A Space Odyssey" and its on-board computer named HAL which spoke and understood just like a human being. Unfortunately it has, for most of the time since then, remained a technology covered in mysticism – a black art – involving such people as "speech scientists" and men with beards and sandals in the fusty corridors of academia. Speech recognition technology falls into three camps – dictation, embedded devices and telephony.
|
|
Telephony provides for speaker independent systems, i.e. any one can use them; large vocabularies, no need for training, and it works over telephone lines from any type of telephone. In the late 1990s, telephony speech recognition started to come of age. High levels of accuracy combined with large vocabularies enabled applications that were not previously possible. One example of such an application is the Odeon FilmLine system developed by Telephonetics. Odeon publishes a single phone number, which covers over 100 cinemas. Speech recognition is then used to let the caller choose the cinema they wish to visit. This would not be a good thing to do with touch-tone but is a perfect application of speech recognition. This system has a success rate of over 98% on the speech recognition part of the application and is used by callers from all parts of the UK encompassing a wide range of accents including Scottish, Welsh, Geordie and Liverpudlian. Other systems deployed in the UK include Lloyds TSB PhoneBank and Barclays Stockbroker system, both of which use speech recognition and both of which are custom developments that have been very successful. However, systems such as these are custom systems involving a good deal of development effort and which are carefully tailored to each customer's requirement. The speech recognition technology is developed by specialist companies who do nothing else. They sell it as a technology component - the speech recognition "engine" – not as a solution in itself. To produce a system such as those mentioned above, a specialist development company like Telephonetics will take all the raw technology components, of which the speech recognition engine is just one, and - with a team of people including linguists, software engineers, hardware engineers and system analysts - will craft a solution. Often, such systems need specialist ongoing help to deal with enhancements and tuning. Tuning for speech recognition is important because the system has to learn about how people react with the system, in the same way that a child learns. The cost of this process means that speech recognition is not prevalent in the same way that auto attendants and voice mail have become. It is not feasible for end users to develop their own solutions unless they have a significant IT department who also possess telephony development skills – not a common scenario! What has been missing is off-the-shelf systems that are affordable and solve real market problems. Such systems need to be proper turnkey, black box solutions that solve real customer issues and which don't need bearded boffins to maintain them. Companies do not want a system that involves a lot of housekeeping or babysitting. They want systems that operate 24 by 7 and with an uptime similar to that which is typical for their PBX. They do not want to become speech recognition gurus. If you visit the web pages of any companies offering speech recognition, you will not find any 'solutions', only technologies or 'platforms'. Happily this situation is now starting to change as a result of companies realizing that actual products that solve business issues are what customers really want.It's a strategy we've adopted at Telephonetics, for example. Our recently launched Voice Dialler (or voice activated directory) fully encompasses this philosophy. One of its features is that it addresses the need which auto-attendants were originally supposed to fill – ie to automate the routing of calls. Touch tone auto attendants never really did the job because people never knew the extension number of the person or department they wanted. Most callers therefore would bale out to the operator. The Voice Dialler has taken speech recognition technology and built it into an actual product that meets a real business need. Now callers can speak the name of a person, department or even job title and the system will understand them and connect the call to who they want. This business solution can form a key part of a company's Customer Relationship Management strategy as it provides a way of answering customers in a consistent professional manner and connecting them straight to who you want and they want to talk to. Such a solution also could be a strategic part of an E-government implementation by offering a 24 7 service to connect tax payers to the right department. From a system administrator viewpoint, this system meets all of the objectives described above. The customer does not need to undertake any lengthy development work or integration work. No speech recognition expertise is required to look after the system. No specialist training is required by any user. Any ambiguities arising from people with the same name are automatically handled by the system. The system works with any brand of PBX and any type of telephone – so can be used by any size of company. One of speech recognition's big opportunities lies in carefully designed business systems which resolve real business needs and do so in an accessible manner so that they can be enjoyed by large numbers of companies, their staff and their customers. Arthur C. Clarke invented the concept of satellites. He predicted space travel, the space station and speech recognition as seen in the film. Speech recognition has come on a long way since the film 2001 and particularly in the last five years. While we have not yet reached Hal's level of fluency and understanding, we can do a lot more than was previously possible - and it won't be too long before HAL's levels of speech recognition become reality. About the Author About the Company |
Published: Tuesday, December 3, 2002
I am checking out all the amazing and daily updated content on ContactCenterWorld.com and networking with professionals worldwide
Send To Friends Post On My Wall