Abstract

This study concerns orthographic standardisation, text normalisation and legacy character encoding as preconditions for machine-readable text corpora that must stand as witnesses in a critical apparatus. It begins from a claim it refutes: that a machine-readable corpus requires a standardised orthography. Swiss German and the Arabic dialects lack one yet support corpora and benchmarks by normalising toward a standard borrowed from a related language (Standard German, Modern Standard Arabic) or toward a convention built for the purpose (CODA). For a corpus that must be authoritative in its own language, a witness to a text rather than a corpus about a language, that borrowing is disqualifying: normalisation that replaces source forms is the inverse of collation and removes the inter-witness differences a critical apparatus records, while a parallel normalisation layer that retains the source form does not. Native orthographic standardisation is therefore argued to be their precondition, with a stated falsification test. The Khmer case, the Pāli canon (Tipiṭaka) translated by a commission of 1929–1968 whose leader also led the codification of Khmer spelling and its dictionary, shows standardisation to be necessary but not sufficient: the orthography was codified by 1967, while character encoding remains unsettled, with five legacy font encodings in one 2009 volume. In the Chaṭṭha Saṅgāyana Tipiṭaka XML, the Cambodian edition has no page-concordance column, and the siglum kaṃ., read as Cambodian, appears in about 8 per cent of variant notes. A further failure of citation-grounded AI follows: a citation that resolves to a rendering, not the source.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS