The Shared Lexicon
Almost everything in Glycerine Aksara belongs to one edition and one editor. Lemmas and inflections do not. They sit in tables shared by every user of the installation, and that single design decision explains most of what is otherwise puzzling about how the Editor treats them.
Why they are shared
Because a lemma is only worth having if it gathers attestations across editions.
If each edition kept its own private word list, the glossary would be a per-edition index and nothing more. Sharing the lemma means the same headword accumulates attestations from every edition that uses it — so the question "where else is this word attested, and how is it inflected there" has an answer, and the answer improves as the corpus grows. The same holds for inflections: a morphological analysis is a claim other editors should be able to reuse and point at.
The cost is that a shared record cannot be freely rewritten by whoever happens to be editing it. Hence everything below.
Private, then canonical
A lemma has an owner, and approval means losing it:
| State | Owner | Who can use it | Who can change it |
|---|---|---|---|
| Private | You | You | You |
| Pending review | You | You | Nobody — it is locked while under review |
| Canonical | Nobody | Everyone | Administrators |
| Deprecated | Nobody | Existing tokens keep it; it cannot be chosen afresh | Administrators |
Why your approved lemma is no longer yours
Becoming canonical is not a flag set on your lemma — it is the removal of its owner. That is the literal mechanism, and it is why the fields grey out the moment a submission is approved. The record has stopped being your draft and become shared infrastructure.
The exchange is deliberate: you give up unilateral control and get, in return, a headword that other editors' work will attach to. A private lemma keeps you editing freely at the price of being a private note. See Lemmas.
Deprecation exists because canonical lemmas cannot simply be deleted — tokens in other people's editions may depend on them. Retiring one keeps every existing link intact while removing it from circulation for new work.
What a lemma actually holds
The headword, a translation, a homograph number distinguishing identical spellings, an optional link to an external dictionary, and a list of cognates — each a (language, form) pair rather than a free-text note, so that the Sanskrit or Pali comparandum is a queryable value.
That is what the structure buys: the Editor searches lemmas by their Skt. cognate, and lists the lemmas that share one as related — the same word under different Gāndhārī spellings, which is otherwise the hardest thing in the lexicon to find. See Related lemmas.
A cognate is not the same thing as a canonical lemma
"Canonical" describes a lemma's review status. The Sanskrit or Pali form beside a Gāndhārī headword is a cognate, and a private lemma can have cognates just as a canonical one can. The two are unrelated.
What an inflection actually holds
Nothing of its own. An inflection is a bundle of values drawn from controlled vocabularies — part of speech, and then whichever of case, number, gender, person, tense, voice, and the rest that part of speech admits.
There is no free-text field, by design: an inflection is meant to be compared and counted across the corpus, which requires that everyone recording an accusative singular records the same value. It is also why the form reshapes itself when you change the part of speech — the visible fields are computed from your choice rather than fixed. See Inflections.
Attestation is per edition, the word is not
A lemma is global; the attestations you see are scoped to what you are reading. The Viewer's glossary lists the lemmas attested in the open edition, with that edition's occurrences — not every occurrence in the corpus.
And an attestation is only ever visible if the edition containing it is. A private edition's tokens contribute nothing to anyone else's view of a shared lemma, so publishing an edition is also what contributes its evidence.
Inherited from READ
The lexicon was seeded from READ, and the sorting algorithm that orders it is a port of READ's own. Editors coming from that environment should find the headwords, the homograph numbering, and the glossary ordering familiar; where they differ, the imported rows are the older authority.
Related
- Lemmas — choosing, creating, and submitting them.
- Canonical Lemmas and Review — the review side.
- Akṣaras and Transliteration — how the ordering is computed.