Speech Industry Must Innovate and Continuously Improve

Speech recognition technology has made strong advances in the past number of years and will likely continue to do so. Recognition error rates are decreasing and additional features and capabilities have been incorporated into the most recent versions of the software.

Progress has also been made to improve voice authentication software, text-to-speech (TTS) software, and statistical model-based natural language speech recognition.

On the human factors side, voice user interface (VUI) designs continue to improve and when combined with the technology improvements mentioned above, are helping speech customers to save money and improve customer service, while enabling callers to enjoy a more productive, automated call experience.

Notwithstanding the technology and VUI design improvements achieved to date, there is more work to be done. More successful deployments and case studies are required to ensure that speech continues to play a strong and expanding role in the customer relationship management ecosystem.

We all recognize the need for continuous improvement in the speech industry, but why the sudden urgency?

Simply put, because enterprises have a host of other tools at their disposal that can take the place of speech to defray call center costs and improve service:

• Touchtone
  The incumbent rival, it is usually cheaper than speech but with functional limitations. For some automated applications today, touchtone remains a powerful, effective tool and the most appropriate self-service mode. Moreover touchtone applications can usually be designed and coded more quickly and affordably than the average speech application today, which makes them an attractive alternative for enterprises.
• Overseas call center agents
  Depending on the application, overseas agents represent a very affordable alternative to speech, despite hefty start-up costs. Overseas call centers provide professional, inexpensive service. While not a perfect solution for every enterprise, overseas agents handle straight-forward inquiries capably and effectively and their services are viewed by some companies as an alternative to speech for certain call types. Outsourced options are also available for companies that prefer not to go it alone.
• Work at-home agents
  Generally more expensive than overseas agents, it still is potentially competitive for speech depending on call type and related factors. Home-based agents can handle all the calls that a typical brick-and-mortar call center agent can, with the added advantage that enterprises don’t have to tie up their capital in physical facilities to house the agents. Overhead is low and agent productivity levels are equivalent to – or better than – those in the traditional call center environment. Plus, there are some affordable outsourced options in the market if enterprises don’t wish to make the investment on their own.

Because touchtone and live agent options are already entrenched in the market, these tools are sometimes incorrectly viewed as competitors to speech. Our challenge is to teach the industry how speech can work in tandem with both, to the benefit of both enterprises and end users. 

So what are the steps the industry should take to ensure that speech recognition continues to be an effective complement to more traditional options?

• Simplify speech design and implementation
  Speech applications need to be easier and faster to design, build and manage. A number of speech ecosystem companies, including those providing platforms, IDEs and packaged applications, are addressing this challenge – and the work they are doing is integral to growing the speech industry and building and deploying great speech apps. Speech customers – and potential speech customers – should explore these products or ask their outsourcer to investigate using them if they have not done so already.
• Continue to improve recognition rates and add features
  Industry leading software vendors need to continue to invest in research and development to improve speech engine performance. In addition, they need to bring additional, value-added products to the marketplace based on their real-world deployment experiences. Both actions will lead to more successful speech implementations and case studies.
• Continue to make speech more caller-friendly
  The industry needs to continue to focus on designing outstanding voice user interfaces (VUIs) for speech applications, incorporating clear dialogs and superior error handling techniques. Additionally, there needs to be more attention paid to application persona. Some companies in the speech ecosystem have done a great job at designing friendly, easy to use applications; others should model those efforts. To the extent callers have an enjoyable, effective, efficient experience with speech, they will use it – otherwise, they’ll opt for an agent.
• Continue to support speech standards; they give customers more deployment alternatives and flexibility
  Thanks to the VoiceXML standard, enterprises now have more speech and touchtone deployment alternatives than ever before: premise-based, hosted, or fully managed (the latter two of which are outsourced). Moreover, they now have the opportunity to work with a growing range of vendors who can collaborate to bring truly “best of breed” solutions to fruition in a standards-based environment. And although standards-based applications are not 100 percent portable yet, they are an improvement over proprietary systems.

Why does it all matter? Because customers who have an expanded list of deployment alternatives are more likely to have the context in which they can envision and generate a strong ROI for speech.

Speech recognition is an effective tool that has saved enterprises millions of dollars while helping them improve customer service. In order for the presence and viability of speech in the customer relationship management arena to continue to grow, speech ecosystem companies need to continue pushing forward with improvements in their respective domains. With continued focus on making speech more effective, easier to implement, and standards-based, the industry will continue to grow and prosper as a viable complement to touchtone and agent-based alternatives.