The sources are TEI P5 XML, normalised to Unicode NFC. Divisions, blocks, editorial metadata and CTS identifiers are kept as they are.
References
A canonical reference (book, chapter, line) is given only where the source attests it. Otherwise the passage keeps a local reference; nothing is guessed.
Presentation
The reader lays out prose, verse, drama, scripture, lexica, commentaries, mathematical texts, fragments and papyri each in their own way. Structural labels are kept apart from the author’s headings.
Words in the reader
Clicking a word shows its lemma, predicted from context by Sôtêr mini, an automatic lemmatiser that keeps its hesitations and gives no calibrated confidence score. The short gloss comes from the Liddell–Scott–Jones lexicon or the abridged Bailly. The morphological analysis comes from Morpheus (Perseus Digital Library), through the data of Eulexis (Philippe Verkerk), under the GNU GPL v3: it describes the form out of context, so every analysis compatible with the lemma is shown. The dialect of a work comes from an editorial table by author; when it is only inferred from the date or the genre, the page says so.
Syntax
In the editions where it is available, clicking a word also shows its analysis in the sentence: its part of speech, its morphology in context, the word it depends on and its function (subject, object, adverbial…), and the words it governs. The whole sentence can be drawn as a dependency tree. This analysis is automatic: it follows the Universal Dependencies guidelines for relations and the morphological tags of the Perseus treebank, and is produced sentence by sentence by a language model guided by Sôtêr’s lemma and the analyses Morpheus allows for each form. Checked against samples of three treebanks (Perseus, PROIEL, PTNK), it gives between 79 and 89% of words their right head, and between 63 and 82% their right head and relation, depending on each treebank’s conventions. It can be wrong. A sentence whose text has changed since it was analysed is not shown.
What is not done
No biography, translation or summary is added. Every page links its metadata, its CTS URN, its digital source when known, and its corpus release.