Vertigo Software created voice-only applications for the retail and financial sectors

Two sample applications based on the Speech Application Language Tags (SALT) specification were released by Microsoft in April. The applications are voice-only applications (that is, they work over standard telephones without requiring any “multimodal” capabilities.) The sample applications were developed by Vertigo Software Inc., a provider of software development and consulting services, using Microsoft’s SALT-based Speech Software Development Kit (SDK) beta 2. Vertigo has developed Web-based ASP.NET sample applications for Microsoft in the past, according to Scott Stanfield, Vertigo CEO. The Speech SDK integrates into the Visual Studio .NET development environment familiar to Web developers.

The sample applications are important in at least four ways:

1.    They provide developers an example of a complete application that they can use to understand SALT and speech-application development.

2.    The specific applications can serve as a template for rapid development of similar applications.

3.    The sample applications illustrate how pre-existing graphical Web resources can be re-used, in particular, the business layer and data layer code. Both sample applications extend an existing Web sample application, and a voice transaction will automatically be reflected in the Web application. James Mastan, director of marketing for Microsoft speech technologies, noted that those specific sample Web applications have been studied by many developers, and provide a familiar base to build upon.

4.    The applications will give developers an insight into the differences in Web-site Graphical User Interface (GUI) design and Voice User Interface (VUI) design. Stanfield said that Vertigo designers were challenged in creating a quality VUI. “It’s still mostly an art,” he noted. “We did lots of testing, and refined the design through trial and error.” He said that Vertigo is developing VUI specialists, reflecting the differences between GUI and VUI designs.

Retail and financial examples

The two sample applications are examples of retail- and financial-sector applications:

§         The speech-enabled ASP.NET Commerce Starter Kit is based on the existing Web sample application “IBuySpy Store,” a tongue-in-cheek commerce site for spies (selling items such as water escape vehicles). Users can use speech recognition to order an item by product number, browse the store catalog, hear product descriptions, and add products to their shopping cart by voice.

§         The speech-enabled Fitch & Mather Stocks (FMStocks) Web application is an online stock brokerage that allows customers to manage a stock portfolio by telephone. Users of FMStocks can obtain quotes on stock prices, buy and sell stock, and review their portfolios.

The applications use directed-dialog style, but with some flexibility and sophistication. For example, the application uses confidence scores to decide whether to verify a choice.

Developers can download the sample applications on Microsoft’s Web site by visiting www.microsoft.com/ speech/ techinfo/ sampleapplications. The web site also includes detailed white papers that contain specifics on the voice interface design and some of the principles involved; these papers may be interesting to those who are interested in looking at an early SALT application without having to set up the development tools.

Xuedong Huang, general manager of Speech Technologies at Microsoft, summarized: “Providing developers with speech-enabling best practices gives them immediate access to this groundbreaking technology, allowing them to produce speech applications that can help reduce costs, enhance revenues, and improve business agility for their businesses or their customers’ businesses.”

Voice platform

The sample applications run on a PC with an attached microphone, using Microsoft’s speech recognition engine and recorded speech. Microsoft is planning a beta release of its Microsoft SALT-based speech platform this summer, and that platform will allow calling the sample applications by phone, Mastan said.

A number of companies are creating SALT-based solutions (SRU # 118, April 2003, p. 1). Microsoft is actively promoting telephone and multimodal solutions with a series of Webcasts and other initiatives. They are also offering prizes for the best “speech controls,” speech components for building applications, in a contest at www.componentsource.com/ SpeechContest.

System using BeVocal hosting at Venetian Hotel, Las Vegas, and elsewhere

On April 7, OnCall Systems, Inc. announced its speech-based telephone recruiting application and the signing of The Venetian Hotel in Las Vegas as a customer. The bilingual (English and Spanish) system is targeted at hourly workers, rather than the salaried employees targeted by most Web recruiting sites. The system is hosted by BeVocal for OnCall using Nuance speech recognition, but Udhe Abluwalia, OnCall president and CEO, said the system is designed for easy portability.

Abluwalia said that the telephone-based, hourly-worker recruiting solution is OnCall’s focus. It was conceived to meet a need, and speech recognition only became a requirement when they realized that tasks such as the caller choosing a job type (with over 600 job titles at the Venetian, for example) could not be addressed satisfactorily with a touch-tone system. Reena Jadhav, OnCall Chief Marketing Officer, said, “Hourly workers, which make up over 70% of U.S. workers, are much less likely to have access to the Internet, but they do have telephones.” (See the article on Internet access in the May issue, p. 17.)

The system is designed to be a continuing resource for workers and employers. When potential employees call, they are given a voice mailbox that they can check for messages or job offers or appointment requests. The system also has outgoing alerts for job offers and interview appointments. The outgoing calls are interactive, for example, to verify that the person answering the phone is the called individual (using a user passcode) and to allow the potential employee to choose a time for an appointment and confirm it on the spot. In addition to providing a tailored response for specific employers, OnCall plans to establish 1-866-JOBFLASH as a multiple-opportunity job line, where workers can be offered a choice of companies and job categories. As workers enter the system, employers can often call up and find a qualified individual immediately, rather than having to post a job requirement. Workers can also call in at any time to update their “resume.” Abluwalia indicated that OnCall’s layered architecture makes the expansion to a job marketplace an easy technical evolution.

When individuals first call in, they are interviewed by the speech recognition system. Some questions, such as availability, job preference, and years of experience are interpreted by the speech recognition system and entered into a database for screening. Answers to other questions are recorded, allowing more lengthy answers, as well as letting employers hear the communication style of the applicant. Some questions are job-specific or employer-specific. An employer can set up screening criteria, and, if a caller meets those criteria, the system can set up an appointment during the initial call. Some applicants are so surprised by the immediate response, Jadhav said, that they call to confirm that the appointment is real. The best applicants go quickly, Jadhav said, so the quick response often gives employers using OnCall an advantage in getting the best employees. Employers can review the database of resumes by phone or through a Web interface.

Dave Newton, vice president of human resources for the Venetian, said, “Our goal is to have our employment people spend their time interviewing the most qualified candidates. OnCall enables us to do this by automatically prescreening and scheduling interviews with qualified candidates for open jobs. It also allows us to collect resumes for future openings.”

The system has already handled over 25,000 calls from about 4,000 individual applicants in three months, according to Jadhav. In addition to the Venetian, OnCall hotel customers include Sheraton Palo Alto, Westin Palo Alto, Marriott San Mateo, and the Hyatt San Jose. With hotel occupancy rates very low in Silicon Valley lately, Jadhav said, OnCall turned to the Las Vegas market to get more traffic on its system. (The Venetian is unveiling a new hotel tower and expanded convention space in June 2003, which increases its existing requirement for new hires.) Notably, however, even though some Silicon Valley hotels are not currently hiring, they are still paying a monthly fee to remain on the system; they save money by not having to manually process and reply to the resumes. The company is also testing the system with security companies, including First Alarm Securities.

OnCall offers the system on a service bureau model with a startup fee and monthly subscription fees. Jadhav said that the monthly fees were chosen to result in an overall cost of about $100 per hired employee. She estimates that this compares to other alternatives as follows: recruiters, $12,500; newspaper, $5,000; job fair, $3,000; campus recruiting, $2,000; and the Internet, $1,000. Some of the efficiencies of the OnCall service include avoiding advertising costs, fewer employees handling resume screening and interviews, a more objective applicant-screening process that reduces job discrimination claims, decreased hiring time, 24/7 operation (particularly important for applicants that are currently working), and an employee database that is more current (because applicants can update it by phone). Jadhav said that there was also the opportunity to reduce employee turnover by correlating the screening criteria with the resulting length of employment of the worker.

Jadhav estimates the available market that the OnCall service addresses at $1l billion, based on the job turnover of about 110 million hourly workers. Competition in the telephone-automated segment is limited to a few companies with narrow touch-tone solutions, she said. (For comparison, the temporary-placement companies have about $60-$120 million in revenues and online recruiting generates about $1-2 billion, she estimates.)