en.wikipedia.org/wiki/Arabic_diacritics
2 corrections found
can only appear on top of an alif (آ)
This is too absolute: maddah is not limited to alif. Unicode’s Arabic-script specification says that while it is usually above alef in modern usage, in Quranic Arabic it also occurs above waw or yeh.
Full reasoning
Unicode’s Arabic-script chapter explicitly says maddah is not restricted to alif in all Arabic-script usage. It notes that in most modern languages using the Arabic script maddah occurs only above alef, but in Quranic Arabic it occurs above waw or yeh to mark vowel elongation. So the claim that maddah “can only appear on top of an alif” is incorrect as a general statement.
A more accurate wording would be something like: in ordinary modern orthography, maddah is usually written over alif, but Quranic orthographies also use maddah above other letters such as waw and yeh.
2 sources
- Chapter 9 – Unicode 17.0.0
“In most modern languages using the Arabic script, it occurs only above alef... In Quranic Arabic, maddah occurs above waw or yeh to note vowel elongation.”
- Chapter 9 – Unicode 16.0.0
“The maddah is a particular case of a harakat mark... In most modern languages using the Arabic script, it occurs only above alef... In Quranic Arabic, maddah occurs above waw or yeh to note vowel elongation.”
For Modern Standard Arabic, the state-of-the-art algorithm has a word error rate (WER) of 4.79%.
This performance figure is outdated. Newer 2024 Arabic diacritization papers report better best-in-class MSA results, including 3.4% WER and a 20% WER reduction over older benchmarks.
Full reasoning
The sentence is written in the present tense (“the state-of-the-art algorithm has... 4.79%”), but later work has already surpassed that number.
A 2024 ACL Anthology paper reports “the lowest reported word-level diacritization error rate for MSA (3.4%)”. Separately, a 2024 arXiv paper reports state-of-the-art results with a 20% reduction in WER compared to existing benchmarks. That means 4.79% is no longer the current state of the art for Modern Standard Arabic.
So the issue is not that 4.79% was never reported—it was—but that the article presents it as the current best result when newer published results are lower.
2 sources
- Arabic Diacritization Using Morphologically Informed Character-Level Model - ACL Anthology
“We achieve the lowest reported word-level diacritization error rate for MSA (3.4%).”
- Arabic Text Diacritization In The Age Of Transfer Learning: Token Classification Is All You Need
“PTCAD ... achieves state-of-the-art results, including a 20% reduction in Word Error Rate (WER) compared to existing benchmarks.”