ATCCaps: A Call-Sign-Aware Speech Dataset for Air Traffic Control Recognition

2026-06-21

AI Summary

This study addresses the scarcity of high-quality annotated data for callsign recognition in air traffic control (ATC) speech by introducing ATCCaps, the first large-scale, callsign-aware dataset. Constructed from real-world ATC recordings, ATCCaps integrates confidence-aware transcriptions, ADS-B-derived callsign metadata, standardized formatting rules, and large language model–assisted generation to deliver multi-level audio-text aligned annotations—including verbatim transcriptions, normalized callsigns, and ATC-style captions. The dataset comprises 202.94 hours of audio, 170,385 utterances, and 922 standardized callsigns, enabling tasks such as callsign matching and audio-text retrieval. Benchmark experiments demonstrate its effectiveness for automatic speech recognition and callsign modeling, while underscoring the necessity of explicitly evaluating callsign and digit fidelity.

Abstract

Call signs are safety-critical entities in air traffic control (ATC) communications because they identify the target aircraft of each spoken instruction. This paper presents ATCCaps, a call-sign-aware ATC speech dataset with caption-level audio-text supervision. Built from real ATC radiotelephony recordings, ATCCaps contains 202.94 hours of curated audio, 170,385 utterances, and 922 unique normalized call signs. The construction pipeline combines confidence-aware transcript parsing, ADS-B-derived call-sign metadata, call-sign normalization, rule-based quality filtering, and LLM-assisted caption generation. Each retained sample is paired with transcript descriptions, call-sign descriptions, and ATC-style captions, supporting ASR evaluation, call-sign matching, and call-sign-aware audio-text retrieval. We further characterize ATCCaps through split statistics, call-sign coverage, seen/unseen call-sign analysis, filtering audits, and caption quality evaluation. The evaluation subset is derived from the human-annotated ATCO2-test-set, enabling reference evaluation with manual transcripts. Results show that ATCCaps provides scalable audio-grounded call-sign supervision, while caption analysis highlights the need to explicitly validate call-sign and numeric fidelity. Reference ASR and CLAP-based baselines demonstrate the usability of ATCCaps for call-sign-aware ATC speech modeling.

Problem

• call sign
• air traffic control
• speech recognition
• audio-text supervision
• safety-critical communication

Innovation

• call-sign-aware
• ATC speech dataset
• audio-text supervision
• LLM-assisted captioning
• ADS-B metadata integration