starlineventures/gulf-coast-ga-atc-communications

墨西哥湾沿岸通用航空ATC通信数据集是一个为墨西哥湾沿岸地区（德克萨斯州、路易斯安那州、密西西比州、阿拉巴马州、佛罗里达州狭长地带）的飞行学员设计的飞行员与空中交通管制（ATC）无线电交流的对话数据集。该数据集旨在微调语言模型，作为航空无线电通信的学习辅助工具。数据集包含5,401个对话，分为训练集（4,860个）和测试集（541个），其中包括3,400个墨西哥湾沿岸合成场景和2,001个真实ATC录音（转录）。数据集涵盖11种以上的场景类型和26个墨西哥湾沿岸机场。数据集有两种配置：一种用于模型微调（sft），另一种用于人类学习和分析（study）。数据来源包括合成场景、布拉格机场的真实ATC语音和捷克机场的ATCO2项目语音。

The Gulf Coast General Aviation ATC Communications Dataset is a conversational dataset of pilot–ATC radio exchanges designed for student pilots training in the Gulf Coast region (Texas, Louisiana, Mississippi, Alabama, Florida panhandle). It is built for fine-tuning language models as study aids for aviation radio communications. The dataset contains 5,401 conversations, split into a training set (4,860) and a test set (541), including 3,400 Gulf Coast synthetic scenarios and 2,001 real ATC recordings (transcribed). It covers over 11 scenario types and 26 Gulf Coast airports. The dataset is available in two configurations: one for model fine-tuning (sft) and another for human study and analysis (study). Data sources include synthetic scenarios, real ATC speech from Prague airport, and ATCO2 project speech from Czech airports.