Design of Arabic Diacritical Marks
Summary
The paper looks at how Arabic diacritical marks should be placed and sized in digital typesetting, and compares this with Latin script. In Latin, a diacritic mostly attaches to one base letter. In Arabic calligraphy, marks also fill the space created by cursive joining, multilevel ligatures and Kashida-based justification. The authors cover the history and classification of diacritics in both scripts, the way Unicode encodes them, and why Latin-oriented strategies and OpenType mechanisms are not enough for Arabic.
Working from Naskh samples by the calligrapher Mohamed Chawki, they found that a mark's position and size depend on the "mass" (height and width) of its base glyph, the following glyph and their height difference. From this they built a phased, left-to-right algorithm. It treats a ligature as a sequence of base glyphs, each carrying its own mark, and stretches Fatha/Fathatan where space needs filling. They implemented it in an OpenType font with TeX justification. The work shows OpenType's limits in substitution and glyph reordering, and argues that Arabic must be handled as a set of distinct calligraphic styles.
Key points
- The paper sorts Arabic diacritics into three groups:
- language marks placed above, below or through a letter (e.g. Fatha, Damma, Soukoun, Shadda, Kasra, Wasl);
- aesthetic marks that fill space;
- explanatory marks that separate Muhmal (dotless) letters from Muajam (dotted) letters.
- Arabic marks are spread over four Unicode blocks: Arabic (0600–06FF), Arabic Supplement (0750–077F), Presentation Forms-A (FB50–FDFF) and Presentation Forms-B (FE70–FEFF).
- Because Unicode does not encode dotted Arabic letters as composites, dots cannot adapt flexibly to multilevel layout and justification.
- The study covers only the Naskh style; the other main styles are Riqaa, Thulut, Maghrebi and Diwani.
- OpenType showed two limits:
- Kashida substitution only works after a good insertion point has been chosen;
- ligatures force a bad choice: either ignore them, which loses multilevel information, or reorder glyphs and marks.
Interesting ideas
- I1. The algorithm places each mark at a default size, then may revisit the previous mark (only for Fatha/Fathatan) once the next glyph is known; this is a bounded one-step look-back instead of a global search. — For Mova (tokens): A tagger or lemmatizer could commit greedily but allow a narrow, rule-gated revision of the previous token's decision, keeping each change traceable.
- I2. A ligature is modelled as several base glyphs, each owning its own diacritic, rather than as one opaque unit. — For Mova (tokens): This mirrors splitting UD multiword tokens (e.g. contractions) into syntactic words, each with its own features, while keeping the surface form intact.
- I3. Positioning should be re-run after justification, because later stages (Kashida, alternate glyphs, ligature toggles) invalidate earlier layout decisions. — For Mova (world): After the world of the text is built, earlier frame or coreference choices that conflict with it could be re-evaluated in an explicit second pass.
- I4. The same mark may be encoded precomposed with its letter or as a separate combining sign, and software must handle both. — For Mova (tokens): The tokenizer should normalize precomposed and combining accents (e.g. in "café" or names) so lemmas and lookups do not split on encoding variants.
- I5. The authors openly state that a two-letter context window is a simplification and that real contexts often span more letters. — For Mova (learn): Evaluation sets could include cases built specifically to exceed a component's assumed context window, which would expose where local heuristics fail.
Cite
@misc{hssini2011design,
title = {{Design of Arabic Diacritical Marks}},
author = {Hssini, Mohamed and Lazrek, Azzeddine},
year = {2011},
doi = {10.48550/arxiv.1107.4734},
eprint = {1107.4734},
archivePrefix = {arXiv},
url = {https://doi.org/10.48550/arxiv.1107.4734}
}References
- B. V. Venkata Krishna Sastry, “Enhanced Font Features for Future Multilingual Digital Typography with Sound-Script-Language Attribute Integration”, TUGboat, Volume 24, Number 3, 2003. find
- M. Solomon, “The art of typography”, Art Direction Book Company, Revised edition, 1995. find
- J. C. Wells, “Orthographic diacritics and multilingual computing”, Language problems & language planning, ISSN, Volume 24, Number 3, 2000, pp. 249-272. find
- J. Victor Gaultney, “Problems of diacritic design for Latin script text faces”, http://www.sil.org/, 2008. ↗
- A. Holkner, “Global multiple objective line breaking”, Honours Thesis, Melbourne, 2006. find
- http://www.typographie.org/, 2009. ↗
- V.Atansiu, “Le phénomène calligraphique à l’époque du sultanat mamluk”, PhD. Thesis, Paris, 2003. find
- M. Hssini, A. Lazrek and Mohamed J. E. Benatia, “Diacritical signs in Arabic e-document”, CSPA08, The 4 th International Conference on Computer Science Practice in Arabic, Doha, Qatar, 2008 (in Arabic). find
- P. Andries, “Introduction à Unicode et à l’ISO 10646”, Document numérique, Volume 6, 2002. find
- G. Richard, “Unicode demystified”, Addison Wesley, 2002. find
- Rachid Zghibi, “Le codage informatique de l'écriture arabe”, Document numérique, Volume 6, 2002. find
- M. S. Mahmoud, “Teach Yourself Arabic calligraphy”, IbnSina, Cairo, 2000. find
- H. Albaghdadi, “Korassat alkhat”, Dar Alqalam, Beirut, 1980. find
- Mohamed Jamal Eddine Benatia, Mohamed Elyaakoubi, Azzeddine Lazrek, “Arabic text justification”, TUGboat, Volume 27, Number 2, pp. 137-146, 2006. find
- P. Smrž and P. Sojka, “Word Hy-phen-a-tion by Neural Networks”, FIMU-RS-96-04, 1996, http://www.fi.muni.cz/reports/files/older/FIMU-RS-96-04.pdf. IJCSI International Journal of Computer Science Issues, Vol. 8, Issue 3, May 2011 ISSN (Online): 1694-0814 www.IJCSI.org ↗
- M. Elyaakoubi, A. Lazrek, “Justify just or just justify”, The Journal of Electronic Publishing, Volume 12, Number 2, 2009. find
- Charlott et Peter Fiell, “Graphic design for the 21 st century”, Taschen, Köln, 2003. find
- M. Breiner, “Diacritics positioning system for digital typography”, http://www.freshpatents.com/, 2009. ↗
- Y. Haralambus, “Fontes et codage”, O’Reilly, Paris, 2004. find
- M. Chawki, “Amchak Chawki”, International commission for the preservation of Islamic culture, Istanbul, 1999. find
- M. Hssini, A. Lazrek, Mohamed Jamal Eddine Benatia, “Arabic digital typography design and connect characters issues”, 5th International Conference on Computer Science Practice in Arabic, pp. 19-31, Rabat, Morocco, 2009 (in Arabic). Mohamed Hssini is a Ph.D. student in Department of Computer Science at Cadi Ayyad University. He is a member of multilingual scientific e-document processing team. His current main research interest is multilingual typography, especially the publishing of Arabic e-documents while observing the strict rules of Arabic calligraphy. Email:m.hssini@ucam.ac.ma. Azzeddine Lazrek is full Professor in Computer Science at Cadi Ayyad University in Marrakesh. He holds a PhD degree in Computer Science from Lorraine Polytechnic National Institute in France since 1988, in addition to a State Doctorate Morocco since 2002. Prof. Lazrek works on Communication multilingual multimedia e-documents in the digital area. His areas of interest include: multimedia information processing and its applications, particularly, electronic publishing, digital typography, Arabic processing and the history of science. He is in charge of the research team Information Systems and Communication Networks and the research group Multilingual scientific e-document processing. He is an Invited Expert at W3C. He leads a multilingual e-document composition project with other international organizations. Email:lazrek@ucam.ac.ma. find
This page in Markdown → https://md.research.elamur.ai/free/en/papers/hssini-2011-design-arabic-diacritical-marks/