Burlington Northern and Santa Fe Railway trains rush along 33,000 miles of track across 28 states and into two Canadian provinces, not unlike their 19th century forebears. But BNSF is now using 21st century voice-recognition technology to turn the railroad’s two-way radio communications and paper reporting process into an integrated real-time reporting and broadcast system.
With help from voice software vendors ScanSoft and InterVoice, BNSF built its own Voice Train Reporting system, using engine cab radios and cell phones as an interface to a speech-enabled, IVR (interactive voice response) system. Conductors report train arrivals, departures, pick-ups, and setouts, and the system automatically turns voice calls into data that can be stored in the company’s DB2 databases running on IBM mainframe computers at its Topeka, Kan., datacenter.
BNSF wanted to move beyond antiquated technology, which required conductors to write their reports or call them in to operators who would fill them out, taking hours to get entered into the company’s datacenter. “We were looking to get more-real time information,” says Shannon McGovern, BNSF’s director of network support systems. “As it was, data would be up to 12 hours old, so we wanted to get it earlier.”
BNSF developed its Voice Train Reporting system based on existing radio networking capabilities. The railway tapped ScanSoft, which provided integrated speech recognition, speaker verification, and text-to-speech functionality in its SpeechWorks software. ScanSoft’s software was able to filter out excessive background noise, an essential ingredient for the success of the project.
The BNSF project illustrates how voice recognition technology can meet unique challenges. “The key distinction [from mainstream speech recognition] is that the software was tuned to characteristics of the radio channel,” notes Bill Meisel, president of TMA Associates, a speech recognition consulting company.
BNSF chose another vendor, InterVoice, to build a platform deploying the ScanSoft speech application. BNSF’s internal technology group completed the application to its own specifications.
“Once the conductor has completed reporting, the IVR application takes the information and turns it into data and enters it into BNSF’s Transportation Support System, with updated reporting that shows, for example, when the train has arrived,” McGovern says. “The information is available to everybody [at BNSF] as well as to our customers.”
By using its own MRAS (Mobile Radio Access System) network, BNSF was able to save money and avoid any conflict with radio traffic critical to the safety of train operations. MRAS is an analog radio interface that is connected to the BNSF telephone network. Cost savings from the project is expected to amount to approximately $4.2 million per year.