Chris Hart, CEO, SRC
Q Tell us a little about SRC's background.
A SRC (formerly know as The Speech Recognition Company) was founded in 1995 as a specialized consultancy for desktop speech recognition solutions. The business was predominantly focused on helping professional services organizations and corporations to successfully implement the leading desktop speech recognition packages of the time (e.g., Dragon, now owned by Scansoft and IBM's ViaVoice), through a combination of customization, training and support.
SRC's management had understood for some time the potential of telephony-based speech technologies so, after substantial city investment in January 2001, we broadened our offering to include solutions for mass-market customer contact using telephony-based speech applications.
Today, SRC leads the U.K. market for high quality hosted telephony-based speech solutions as well as continuing to offer a full breadth of speech services and solutions for the desktop market. Services include consultancy, application development, deployment, application hosting and support. This unique offering has enabled SRC to provide some of the most high-profile speech services deployed in the U.K. so far.
With the common mission to deliver cost savings and productivity improvements through turning speech technologies into solutions, SRC's business is directly focussed on helping our customers' business; a strategy that will drive both our business and the growth of the speech industry as a whole.
Q Please describe a successful deployment of speech technology with appropriate supporting statistics.
A Through our experience of delivering speech solutions we have seen that there are a number of key elements that are essential in achieving success. SRC's service for Freeview (owned by BBC, BskyB and CrownCastle) is a good illustration.
Freeview, Britain's new free-to-view digital terrestrial television service, provides 30 TV channels plus interactive services and digital radio through a rooftop aerial. SRC designed, deployed and operates an automated telephone service for Freeview allowing callers to determine whether or not their postcode area has signal coverage and to request a Freeview information pack.
SRC combines a number of key technologies in the Freeview service, including Nuance's speech recognition engine and Rhetorical's rVoice text-to-speech engine. These technologies are applied in SRC's ContactCapture™ application which handles the name and address elements of the Freeview service, and is hosted on SRC's carrier-grade speech platform which is built on Nortel Networks' media processing platform.
Our experience in designing and deploying these kind of services enabled SRC to develop and bring live the Freeview service in just a few weeks.
The service for Freeview has proved extremely successful, with nearly one million calls handled and, most crucially, more than 5000 calls being handled in the minutes following each TV trailer. This is an example of a speech recognition service being used where a call-center-only solution would have been impractical because it wouldn't have been able to handle the massive peaks in calls generated by the prime time advertising slots.
Q What is your outlook concerning speech technology? Please discuss providing your thoughts in both short and long perspective.
A The technology has advanced to the point where sophisticated and cost effective speech-based interfaces can be developed and deployed, the possibilities for speech-based technology are therefore substantial. However, the challenges remain in selecting the right opportunities for the application of speech technology and then implementing in a way which will deliver satisfaction to the end-user. Short term, the market success will be driven by suppliers' ability to tackle these two challenges. In the medium term, we will see further improvements in speech recognition (ASR) and some rapid developments in text-to-speech (TTS) and speaker verification resulting in more widespread commercial deployment. In the longer term, advances in ASR and associated technologies will lead to a move towards true "natural language" speech recognition, opening up the range of application which speech can successful tackle.
Q What vertical markets are the strongest drivers/implementers of the services speech technology has to offer?
A We've been in discussions with companies from a diverse cross section of markets and found that the common driver for speech services is not so much industry-based as process-based. The driver can be identified in any function that involves high-volume customer contact and repetitive information transactions. It's this diversity that makes our business exciting.
That said, there are several vertical markets that have been quick to recognize the benefits that speech services can deliver and obviously reference deployments remove some of the barrier to rapid uptake of speech by other companies in that market. Telephone betting has seen a lot of activity, with SRC providing solutions for a number of U.K. bookmaking companies. We've also seeing a high concentration of activity in the Financial Services sector, companies like Lloyds TSB deploying paperless share registration services in addition to their telephone speech banking service.
Q Please explain how human factors play a role in speech applications.
A I'm glad you asked this, because I can't stress enough how important the perceived usability of automated speech services is to their success. It sounds like an obvious concept, but we've seen services launched that aren't intuitive and insist on rigidly following principles established by touch-tone systems.
Time invested in specifying the functionality of the service, involving end-users throughout the development process and usability testing is essential in ensuring the success of the speech service. The reaction of callers to services is not always predictable and user testing and development iterations are vital. SRC places considerable emphasis on the role of human factors from project outset. The beginning of the process involves observing current users behaviour and gathering information about the service usage, demands, goals and rationale. We've also found that the choice of persona can dramatically influence the 'stickiness' of a service. The more the voice of the speech service has the right balance of friendliness and professionalism coupled with the right tone and pace, the more effective it is at taking care of customers and the more positive their experience with the service.
Ultimately, incorporating human factors into the development process enables SRC to ensure user acceptance of our speech services when they are brought into volume service.
It's important to remember that at this stage, callers' experience of automated speech applications is fairly limited and a negative experience of a service can do little but harm the speech industry as a whole. As speech solutions providers we have a responsibility to 'fast-track' acceptance by making usability a key consideration for our applications.
Q Who are your partners and why did you choose these partners? How do you decide to partner with a company?
A SRC's strategy has always been to work with best-of-breed suppliers, to that end we have forged relationships with a number of organizations that are at the forefront of their respective technologies. In the speech recognition segment our partnerships include Nuance, SpeechWorks and ScanSoft, for text-to-speech we currently work with UK-based Rhetorical, and for our hosting platforms we have a very productive relationship with Nortel Networks whose technology for media processing and call center environments is core to our speech application hosting services.
We've also developed close marketing relationships with a number of leading outsourced service suppliers who have recognized the benefits of providing high-quality automated speech services as part of their offering.
Q What do you hear from your customers concerning speech technology?
A From existing customers the most common comments are surprise at how well the speech services are working and interest in exploring other areas where they can utilize speech to improve business effectiveness.
When approaching new customers, the initial challenge SRC experiences is overcoming scepticism surrounding the technology and likely user reaction. For some, speech recognition has been tarnished by bad experiences of using poorly designed touch-tone IVR systems or early attempts at speech services which left a lot to be desired in the usability stakes. It's these misconceptions that can lead people to believe that speech recognition is not ready for serious commercial use.
Of course, one of the pleasures of our business is in showing customers just how effective speech services can be and it's always gratifying to hear how impressed they are after using a well designed speech service.
Once this hurdle has been overcome customers are quick to recognize the massive potential for introducing an automated speech service and soon begin to identify processes in their organization that can benefit not only in terms of saving costs and increasing productivity, but also by improving customer service and maintaining consistent branding.
Q What should the speech technology industry as a whole be doing to increase the growth rate of speech technology deployments?
A As I mentioned earlier, a focus on usability and human experience must be key to any player in the automated speech services market. In fact, as a growing industry, we all have a responsibility to ensure our services offer the very best user experience possible, for it is this misconception that has hampered the uptake of speech services to date.
Many companies have been running pilots and trials for a long time, but never manage rollout to high volume commercial usage because the services don't live up to their performance promises - and this is mostly due to poor usability. The automated speech services market needs to establish more high profile successful volume deployments. Lloyds TSB's Phonebank Express and the information enquiry service supporting BBC Freeview are a good start. Once callers realize that speech automated services can be so much better than IVR touch tone, then mass acceptance of speech technology solutions will follow. Uses such as voice portals, voice commerce and enterprise applications also represent substantial future market opportunities.
Q What are your thoughts concerning standards and their impact on speech technology?
A The so-called "current standards war" between VoiceXML and SALT doesn't really exist as they each aim at different areas of speech service development. However, the latest "V+X" standard initiative is to be encouraged since it should lead to convergence. Standards will certainly help the speech industry establish itself in the mainstream.
VoiceXML has already establishing itself as an effective standard for scripting telephony based speech recognition applications in call centers and, as a member of the VoiceXML Forum, SRC is strongly embracing VXML in much of our work.