Research Program
WDYW — Well-formed Digital
Yorùbá Writing
Every WDYW module is released, discussed, and improved in the open.
Read a module, then join the conversation below it — questions,
corrections, and disagreements are all welcome.
Tone Restoration for Standard Yorùbá
The foundational module: a corpus-driven system for automatic Yorùbá
tone restoration, evaluated against an independently-verified
48-sentence gold standard. Introduces the four-layer matching
architecture and a grammar-grounded monosyllabic disambiguation
layer. This is the module behind the live ToneMarker tool.
Try ToneMarker →
Lexical / Spelling Correction
Extends WDYW.01 with fuzzy matching against the verified corpus, so
words absent from it can be corrected toward the nearest attested
standard form — not just toned, but spelled correctly. Expands the
target corpus beyond the tool's current sources to the wider body of
digital Yorùbá text.
Discussion opens once WDYW.02 is published.
Grammatical Correction & Neural Corpus Work
A longer-term research direction: sentence-level grammatical
well-formedness, and the large-scale verified corpus generation
needed before neural approaches become viable for Yorùbá.
Discussion opens once WDYW.03 is published.
