Peer-reviewed research
Diacritic Recognition Performance in Arabic ASR
Hanan Aldarmaki and Ahmad Ghannam, Interspeech 2023, 2023
Read the originalRead on October 17, 2026
- What it is
- A 2023 study from MBZUAI that fine-tuned speech recognition models on Arabic transcripts with and without diacritics, and measured how well the models recovered the short-vowel marks. It explains why most Arabic speech models output undiacritized text.
- How much weight it carries
- Peer-reviewed conference paper from a speech research group. It studies diacritic recognition on specific training sets, so its numbers do not describe every commercial dictation product.
Facts taken from this source
- The 2023 Interspeech paper found that diacritizing inside the speech model outperformed diacritizing the text afterward, especially when the model was fine-tuned on manually diacritized transcripts.
“ASR diacritization significantly outperforms text-based diacritization, particularly when the ASR model is fine-tuned with manually diacritized transcripts” (Abstract)
- The 2023 Interspeech paper says that most people write and type Arabic text without diacritics.
“most people write and type Arabic text without the inclusion of diacritics” (Introduction)
- The 2023 Interspeech paper says that most speech corpora and Arabic ASR models are undiacritized, because diacritics are often omitted.
“so most speech corpora and ASR models are undiacritized” (Abstract)