In this series of posts, we will be discussing the practical application of automatic speech recognition (ASR). Initially, our target domains will be human-in-the-loop simulations and command & control applications.
The target audience is the buyers, managers, and developers of speech-enabled applications. Speech recognition has been a successful technology addition to command & control systems. For instance, ASR can be found in the cockpits of military and civil aircraft. Additionally, speech recognition is used to improve realism and to reduce operating costs in training simulators. For example air traffic control simulation, flight simulation, and military tasks such as close air support and call for fire.
However, even though the benefits of speech recognition in these applications are clear, there is still a reluctance by many to invest in the technology. Costly examples of failed projects are widespread in the industry. Additionally, failures in the implementation of speech recognition are likely to continue for as long as they are poorly specified, and the practical applications misunderstood. For instance, requirements for ASR in simulation and C&C are often summed up in two key sentences; the system must be capable of an accuracy of x%; the system must support terminology as defined in “insert document name here.” Unfortunately, neither of these requirements are consistently measurable between vendors (what is the definition of accuracy?) If you are not more specific about phrase support, how do you know you are comparing multiple solutions on a like for like basis?.
Over the coming weeks, Verbyx will assist you in better understanding the application of speech recognition in these and similar domains. Our team has many decades of domain experience. We will guide how best you might define and measure your requirements to ensure you are comparing like for like.
Finally, we are always at your disposal for a private consultation. Enjoy!