Speech Recognition Hits the Street

There are few communication environments where speed, accuracy and precision are as important and difficult as live equities trading. The floor of a stock exchange is an atmosphere of barely controlled chaos where brokers need to communicate their clients’ needs in an emotionally charged, very noisy environment. It is an atmosphere where the bulls are stampeding, or the bears are growling, and there never seems to be a calm middle ground. The financial services industry also has its own unique language, and with IPOs (initial public offerings) being put forward every day, the vocabulary of Wall Street is constantly changing. However, despite the difficulties, some companies are enjoying success within the financial services market. Despite the difficulties of using speech on Wall Street, the financial services industry still attracts the interest of speech companies. Ficomp Systems Inc., for example, a systems integrator based in Dayton, NJ provides a speech recognition system currently in use at Bear Stearns in New York City, the Bank of Montreal’s Chicago based foreign exchange trading floor and other brokerage houses. Using speaker dependent voice recognition systems, a trader enters an order simply by describing it, bypassing traditional order input systems, such as the keyboard, or passing tickets to clerks. “The productivity of the FX (foreign exchange) floor has increased dramatically since we introduced a speech recognition system,” said Donald Lloyd, manager of Foreign Exchange Trading at the Bank of Montreal’s Chicago office. The Bank of Montreal uses the Interpreter 6000 from FiComp, a data input systems specifically designed for financial applications, which can seamlessly integrate with most existing technology in the financial services industry. Gary Richman, president of FiComp notes that the company’s major challenges in preparing a product that would overcome long-standing resistance on the part of traders and faced several unique challenges in pursuing the market. “The grammars are dynamic,” he noted. “New issues are coming out literally every day. And a speech recognition system has to be able to accommodate that.” Richman also noted there were other challenges in reaching the financial services market. “Overcoming the noise is a huge issue,” Richman said. “And the fact is that no one wants to modify their software until they are absolutely certain the product will work. You have to get the product up and running without changing the application software. The traders don’t want to know about technology. They just want to know the product works and will help them do their jobs.” Richman feels the Interpreter 6000 solves the noise issues up to ninety-five decibels and points out that at the Bank of Montreal used the system for sixty days before they used it live. Since then, they have reported 30% increase in the amount of trades a person could make, with fewer mistakes, Richman said.