Speech Technology Magazine

 

The 2019 State of Intelligent Virtual Assistants

Accuracy rates rise and quality improves
By Paul Korzeniowski - Posted Feb 4, 2019
« previous Page2 of 2
Bookmark and Share

Vendors are developing versions of their solutions for vertical markets as well. In November 2018, IPsoft announced 1Bank, a conversational banking solution powered by its Amelia IVA that allows customers to engage with their financial institutions through voice or chat. Amelia helps resolve complex customer questions using comprehensive dialogue. For example, with the query “How much can I afford to pay for a new home?” Amelia will follow up with important questions such as “What is your annual salary?” or “How much of a down payment would you like to make?”

In September 2018, Nuance worked with Epic to enhance the clinical documentation capabilities of its Dragon Medical solution. The two companies built voice-enabled workflows through Epic Rover, enabling the more than 100 healthcare organizations that already have access to Dragon Medical through Epic Haiku—and the thousands of physicians that use it every day to create voice-driven clinical documentation—to conversationally retrieve schedules and look up patient information, laboratory results, medication lists, and visit summaries.

A Look Ahead

While a great deal of progress has been made, more work remains. “There still are instances when the IVA does not offer the user much assistance,” Miller says. Vendors are trying to address this problem by collecting more use cases and honing the algorithms that help the user. 

These systems rely heavily on artificial intelligence and machine learning. With AI, a person programs an application so it performs certain functions automatically; with machine learning, the application programs itself. The potential with these features is enormous because companies can offload more functions from staff to machines. 

However, building and tuning the applications is complex and time-consuming for a few reasons. With natural language processing, corporations must collect large conversation sample pools and train them to recognize words. The more data collected, the more accurate the models typically are. However, collecting and training is a time-consuming, processor-intensive task. The customer must ensure that sufficient processing resources are available to create and continuously enhance the model.

IVAs take that process one step further. Rather than just recognize words, these systems connect words to actions. Consequently, processing requirements become even more complex, and the system demands grow. This type of application is different from traditional IT systems, as once the latter were installed, they typically required little ongoing maintenance.

Finally, the variety of data sources is increasing. “Companies are pulling information from a variety of databases that are all running independently of one another,” noted Opus Research’s Miller. Companies need new tools to help them manage such processes. 

The bottom line is that intelligent virtual assistants are gaining traction because they are becoming more functional. One ripple effect is the systems are becoming more complicated, so managing them can be more difficult. “Now that companies are deploying IVAs in growing numbers, they will begin to turn their attention to managing them more efficiently,” says Miller. Solutions to address emerging management problems are expected to start to arrive in the coming year. 

Paul Korzeniowski is a freelance writer who specializes in technology issues. He has been covering speech technology issues for more than two decades, is based in Sudbury, Mass., and can be reached at paulkorzen@aol.com or on Twitter @PaulKorzeniowski.

Learn more about the companies mentioned in this article in the Speech Technology Buyer's Guide:
Learn more about the companies mentioned in this article in the Vertical Markets Guide: