InterSpeech 2021

Boosting of contextual information in ASR for air-traffic call-sign recognition
(Oral presentation)

Martin Kocour (Brno University of Technology, Czech Republic), Karel Veselý (Brno University of Technology, Czech Republic), Alexander Blatt (Universität des Saarlandes, Germany), Juan Zuluaga Gomez (Idiap Research Institute, Switzerland), Igor Szöke (Brno University of Technology, Czech Republic), Jan Černocký (Brno University of Technology, Czech Republic), Dietrich Klakow (Universität des Saarlandes, Germany), Petr Motlicek (Idiap Research Institute, Switzerland)
Contextual adaptation of ASR can be very beneficial for multi-accent and often noisy Air-Traffic Control (ATC) speech. Our focus is call-sign recognition, which can be used to track conversations of ATC operators with individual airplanes. We developed a two-stage boosting strategy, consisting of HCLG boosting and Lattice boosting. Both are implemented as WFST compositions and the contextual information is specific to each utterance. In HCLG boosting we give score discounts to individual words, while in Lattice boosting the score discounts are given to word sequences. The context data have origin in surveillance database of OpenSky Network. From this, we obtain lists of call-signs that are made more likely to appear in the best hypothesis of ASR. This also improves the accuracy of the NLU module that recognizes the call-signs from the best hypothesis of ASR. As part of ATCO² project, we collected liveatc test set2. The boosting of call-signs leads to 4.7% absolute WER improvement and 27.1% absolute increase of Call-Sign recognition Accuracy (CSA). Our best result of 82.9% CSA is quite good, given that the data is noisy, and WER 28.4% is relatively high. We believe there is still room for improvement.