Omega becomes a texteme processor - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2005

Omega becomes a texteme processor

Résumé

The distinction between "characters" and "glyphs" is a rather new issue in computing, although the problem is as old as humanity: our species turns out to be a writing one because, amongst other things, our brain is able to interpret images as symbols be-longing to a given writing system. Computers deal with text in a more abstract way. When we agree that, in computing, all possible "capital A" letters are represented by the number 65, then we cut short all information on how a given instance of capital let-ter A is drawn. Modern computing jargon describes this process as "going from glyphs to characters." If a glyph is the image of a writing system's atomic unit, a character is an interpretation of that image, an interpretation shared by many glyphs drawn by different people in different places at different times. If all these drawings are equivalent in terms of in-terpretation, we can consider character as an equiv-alence class of glyphs. To be operational such an equivalence class must be described in a clear and unambiguous way. This is why we define charac-ter as being a description of an equivalence class of glyphs [7, pp. 53-58], [6]. Arabic text provides a typical illustration ground for the concepts of character and glyph. In Arabic alphabet, letters are contextual, in the sense that a given letter will change form according to the presence or absence of other surrounding ones. When we refer to an Arabic letter and represent it graphically, we use the isolated form. We can also refer to it by its description (for example: arabic letter jeem) and this can be considered as de-scription of a "character": the equivalence class of shapes this letter can take in millions of Arabic doc-uments. While there may be millions of instances of this letter, according to Arabic grammar they all be-long to one of only four forms: isolated , initial , medial , or final . Hence, we could choose to have not one but four equivalence classes of shapes: ara-bic initial letter jeem, arabic medial letter jeem, and so on. But are these "characters"? Answering to this question requires a pragmatic approach. What difference will it make if we have one or rather four characters for letter jeem? There will indeed be a difference in operations such as searching, indexing, etc. A good question to ask is: "when I'm searching in an Arabic document, am I looking for specific forms of letters?" Most of the time, the answer is negative. 1 Form-independent searching will, most of the times, produce better re-sults and this implies that having a single character for all forms is probably a better choice. 2 Unicode is a character encoding. In other words, it contains descriptions of characters and tries hard to define characters properly by avoiding dependence on glyphs. 3 1 Arabic words are not always visually segmented as En-glish ones-there is, for example, no guarantee that the first letter of a word will always be in initial form: if a word start-ing with jeem is preceded by the definite article al, then the jeem will end up being in medial form. 2 Greek is different: sigma doesn't "become" final because it "happens" to be at the end of a word. While medial sigma can appear anywhere, final sigma is used mainly for the end-ings of particular grammatical forms and in onomatopeias or foreign words. One would hardly ever search for both the fi-nal and medial form of sigma since their rôles are distinct. To illustrate this, when we abreviate a word by a period at a sigma then the latter does remain medial despite being the final letter: φιλοσοφα → φιλοσ. Hence it is quite logical to use distinct characters for medial and final sigma. 3 This is not always the case because of Unicode's tenth founding principle, namely convertibility of legacy encodings-and legacy encodings contain all kinds of things. For example, again in the case of Arabic, the main Unicode Arabic table indeed contains only form-independent "char-acters." But, hidden towards the end of the first Unicode plane, one finds several hundreds of codepoints containing Arabic letters and ligatures in fixed forms, for legacy reasons. Like human history (or Stephen King's movies) Unicode has shadowy places which people try to avoid and even to forget that they exist.
Fichier principal
Vignette du fichier
hal-02169904.pdf (226.39 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-02169904 , version 1 (01-02-2023)

Identifiants

  • HAL Id : hal-02169904 , version 1

Citer

Yannis Haralambous, Gabor Bella. Omega becomes a texteme processor. EuroTEX 2005, Pont-à-Mousson, 2005, Mar 2005, Pont-À-Mousson, France. ⟨hal-02169904⟩
29 Consultations
12 Téléchargements

Partager

Gmail Facebook X LinkedIn More