Automated Interpretation of Air Traffic Control Communication: The Journey from Spoken Words to a Deeper Understanding of the Meaning

Abstract

Sophisticated Automatic Speech Recognition (ASR) technologies have become increasingly popular and are widely used in all domains over the years. Systems like Google Assistant, Siri®, Alexa® are integrated into our day-to-day lives. These systems offer a wide range of possible applications just by understanding human speech. However, in the Air Traffic Control (ATC) domain, even the most advanced simulators can just partially replace expensive pseudo-pilots. In spite of having a standardized ATC phraseology, it is still a major challenge to recognize and correctly understand the communication between air traffic controllers (ATCo) and pilots. This is because understanding an ATCo-pilot communication requires more than just transforming speech to a sequence of words. For most ATC applications, perfectly recognizing the sequence of words would not be useful, if the meaning behind the word sequence cannot be correctly interpreted. Recently, 20 European partners from Air Traffic Management (ATM) domain have agreed on a common set of rules, i.e., an ontology on how to transform the spoken words into ATC instructions that clearly define the meaning of the words and make them usable for different applications. In this paper, we present an extension of the mentioned ontology to make it usable for pilot speech as well. We also show some of the challenges faced in understanding the meaning of ATCo-pilot communication and describe our approach of tackling them. Furthermore, we present an algorithm to transform words automatically into ontology instructions and describe the interfaces used to ensure a consistent and reliable communication of ATC instructions. This interface includes, besides other information, plausibility values, different speakers, and ambiguous outputs.