README.md

metadata

language:
  - en
task_categories:
  - automatic-speech-recognition
tags:
  - aviation
  - ATIS
  - whisper
  - speech
size_categories:
  - n<1K

T-ATIS Dataset

Dataset Description

This dataset contains aviation ATIS (Automatic Terminal Information Service) recordings for fine-tuning Whisper ASR models.

Dataset Summary

• Language: English
• Domain: Aviation/ATC (Air Traffic Control)
• Task: Automatic Speech Recognition
• Total Samples: 20
• Train Samples: 18
• Validation Samples: 2

Data Fields

• audio: Audio file sampled at 16kHz
• sentence: Human-corrected transcription of the audio

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("Mrjadkowski/t-atis-whisper")

# Access samples
print(dataset["train"][0])

Fine-tuning Whisper

This dataset is formatted for direct use with Whisper fine-tuning scripts:

from transformers import WhisperProcessor, WhisperForConditionalGeneration
from datasets import load_dataset

dataset = load_dataset("Mrjadkowski/t-atis-whisper")
processor = WhisperProcessor.from_pretrained("openai/whisper-base")
model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-base")

# Your training code here

Data Collection

Audio recordings collected from live ATIS broadcasts at various airports. Transcriptions manually corrected for accuracy.

License

Please ensure compliance with aviation data usage regulations.