Digital lexica and reproducible notes

Structured entries

Digital lexica can represent a headword, variant form, grammatical label, nested sense and example citation as separate fields. The TEI Guidelines show how that structure supports precise search and reuse. A plain scan of a printed dictionary is valuable but does not automatically supply searchable lemmata or sense identifiers. OCR can mistake similar letters, punctuation and diacritics; inspect the scan when a search result seems anomalous.

Stable references

Where available, cite a stable entry identifier and a corpus passage identifier, not only a search-results URL. The TLA documents stable lemma URLs; Oracc’s ePSD2 maintains structured glossaries and linked texts. Record a date or release version because lemmatization and readings can change as projects are corrected. A reproducible note includes the attested token and location, proposed lemma, exact sense or subsection, relevant corpus examples and the editorial status of uncertain letters.

If a dictionary display combines a public-domain text with licensed modern content, check the rights for the particular layer you used. Logeion explicitly distinguishes old public-domain references from modern dictionaries shown by agreement. A citation is always appropriate, but citation alone does not grant permission to copy a dataset or page image. This course summarizes method and directs readers to the source interfaces.

A transparent lexical decision

Keep the chain short enough to audit: witness or edition → token → normalization, if any → morphological candidates → selected lemma → candidate senses → cited comparanda → contextual translation. Beside each arrow, identify who made the decision and how certain it is. A parser supplies candidates; a lexicon proposes a sense grouping; the reader argues for this occurrence. If a step is unsupported, mark it as unresolved instead of making a polished translation appear inevitable.

Practice

Create a one-page lexical record for a difficult word. Include exact passage citation, image or edition link, token, alternate reading, lemma candidates, grammar check, dictionary edition and sense, two contextual comparanda, final rendering and one remaining uncertainty. Another reader should be able to repeat every lookup without asking you what search you used.