KhateebAI/Khateeb_audio_44KH_1_27

Dataset Card

Overview

This dataset contains manually reviewed audio samples and their corresponding transcriptions and phoneme sequences.
It is designed for tasks in speech recognition, phoneme alignment, and analysis of pronunciation vs. orthographic transcription.

Features

Column name	Description
audio	The raw audio signal, sampled at 44.1 kHz.
auto_transcription	The text output generated by an automatic speech recognition (ASR) system before manual correction.
correct_transcription	The manually reviewed and corrected transcription of the audio.
auto_phoneme	The phoneme sequence automatically generated from the ASR output.
correct_phoneme	The ideal phoneme sequence obtained by converting correct_transcription to phonemes according to standard rules. For example, in Arabic, a final consonant may end with sukun if it is at the end of the utterance.
sound_matching_phoneme	The phoneme sequence that matches what is actually pronounced in the audio, even if it differs from the orthographic norm. For example, an Arabic word that is orthographically ended with sukun may be pronounced with a final vowel (haraka) in natural speech.
file_name	The original file name or identifier for the audio clip.
phonemes_BW	The phonem representation in BW using Nawar Alhalaby

Dataset Statistics

• Total audio length: 7,367.31 seconds (~2.05 hours)
• Total sound_matching_phoneme character count: 129,951
• Total sound_matching_phoneme word count: 14,397
• Total correct_transcription character count: 129,329
• Total correct_transcription word count: 13,762

Notes

• Manual Review: correct_transcription and correct_phoneme were fully reviewed and corrected by human annotators.
• Usage: Ideal for evaluating ASR output quality, studying phonetic variation, and training models that account for differences between expected and actual pronunciation.
• Language-specific note: In Arabic, differences between correct_phoneme and sound_matching_phoneme often arise due to sentence-final pauses, vowel insertion, or pronunciation habits.

---