Peer-reviewed research
Novel approaches to Arabic speech recognition
Katrin Kirchhoff et al., IEEE Workshop on Automatic Speech Recognition and Understanding, 2003
Read the originalRead on October 16, 2026
- What it is
- A workshop paper from the 2002 Johns Hopkins summer workshop on recognizing dialectal Arabic, which reviews earlier results on the CallHome Egyptian Arabic task.
- How much weight it carries
- Peer-reviewed but old: the systems it reports are from the 1990s and early 2000s, far weaker than current ones. It shows the shape of the dialect problem, not today's error rates.
Facts taken from this source
- On the 1996/97 NIST CallHome task for Egyptian Colloquial Arabic, the best word error rate obtained was 61 percent.
“The best performance obtained in those evaluations was 61% word error rate (WER)” (Section 1)