UWB-ATCC: Air Traffic Controller and Pilot Speech Recordings

The UWB-ATCC Corpus contains 20 hours of recorded communication between air traffic controllers and pilots. The speech is manually transcribed and labeled with speaker role (pilot or controller). It is provided by the University of West Bohemia, Department of Cybernetics, and was last updated on the platform in April 2023.

Use Cases

• Training automatic speech recognition (ASR) models based on the transcribed audio recordings.
• Developing speaker role classification systems based on the pilot/controller labels.
• Studying aviation communication patterns and phraseology based on the transcribed dialogues.
• Creating synthetic or augmented training data for aviation NLP tasks based on the existing corpus.

Strengths

• Contains manually transcribed speech, which suggests high transcription accuracy.
• Includes speaker role labels (pilot/controller), providing a key feature for analysis.
• Provided by an academic institution (University of West Bohemia), suggesting research-oriented curation.

Limitations

• The corpus is described as 'currently small (20 hours)', which may limit model training.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2023-04-04 08:14:06; freshness should be verified for current projects.

Provenance

Source

University of West Bohemia, Department of Cybernetics

Collection Method

Recordings of real air traffic control communication, manually transcribed and labeled.

Time Range

null

Freshness

2023-04-04

Geography

null