Akṣaras and Transliteration

This page is about how Glycerine Aksara models the script, not about the script itself. The model departs from ordinary editorial practice in a few places, and those places are where mistakes get made.

What the system calls an akṣara

In Glycerine Aksara an akṣara is one outlined region of the image together with an ordered sequence of graphemes. That is the whole definition. It carries no notion of a syllable, no onset/nucleus structure, and no rule about what may or may not go inside it.

An akṣara is a sequence of graphemes, not a syllable

śu is not an entry anywhere. It is the akṣara ś + u — two graphemes recorded in order. The catalogue contains only individual graphemes, so searching it for sa, ka or śu returns nothing. See Akṣaras and Graphemes.

This matters beyond the picker. Because the unit is the grapheme, everything that describes the writing attaches at grapheme level too — a text-critical mark applies to the u rather than to the sign as a whole, which is what lets an editor restore a vowel while leaving the consonant as read.

Where the akṣara boundary actually falls

The outline says where a sign is. It does not say where a word is: word boundaries are recorded separately, on the token layer, and they are free to fall inside an akṣara. A sign that closes one word and opens the next is recorded once, and two tokens each claim part of it.

So the editorial decision "where does this word end" never forces a decision about how to outline the writing. The two questions stay separate, which is the point.

The vowel carrier

The one piece of machinery in the model with no counterpart on the page.

A vowel that opens an akṣara, or that follows another vowel, is stored with a preceding carrier grapheme, written ʔ and typed as a consonant. It exists so that every vowel in the data has a consonant before it, which keeps sorting and comparison uniform.

You will not meet it directly. It is hidden from the grapheme picker, inserted automatically on save, and stripped again on display — a + ʔ + i renders as aï. The only reason to know it exists is that it explains the diaeresis, which otherwise looks like something you did not type.

Sorting follows the script, not the Latin

Every grapheme carries a numeric sort code, and those codes encode Indic alphabetical order — vowels, then velars, palatals, retroflexes, dentals, labials, semivowels, sibilants, h. Variant forms sort immediately after the base they modify, so ḵ follows k and ḏ follows d rather than landing wherever Unicode would put them.

Lemmas are sorted by the same codes, computed from the headword. Two consequences worth knowing:

• Editorial markup is stripped before sorting, so a restored word files under its letters, not under its brackets.
• Anusvāra before a stop sorts as the corresponding class nasal, so aṃta files where anta does.

This is why the glossary and the paleography view are ordered the way they are. They are not sorted alphabetically by the transliteration string, and asking why ṣ precedes s has the same answer as it would in any Indic dictionary.

Inherited from READ

The sorting algorithm is a port of the one used by READ, and the shared lexicon was seeded from it. If lemma ordering here matches what you are used to from READ, that is why.

Editorial signs are graphemes too

The signs an editor uses for what is not legible — the fragment break, the lost akṣara, the illegible akṣara, the unread vowel — are entries in the grapheme catalogue like any other, each with a type and a sort position. They are recorded as part of the reading rather than as annotations on it, which means a lacuna occupies a position in the text and can be outlined on the image exactly as a legible sign can.

What is fixed and what is yours

The grapheme catalogue is reference data. It is the same for every user, and editors do not add to it — an akṣara is spelled from the signs that exist, not from ad-hoc strings. The same is true of the vocabularies for clarity, obscuration, text-critical marks, and grammar.

What is yours is the reading: which akṣaras exist, where they sit, which graphemes each is made of, and how they are marked.

Related

• Akṣaras and Graphemes — recording them in the Editor.
• Text-Critical Marks — the six marks and their brackets.
• Reading the Transliteration — diplomatic, hybrid, and reconstructed.