A computational system for ancient Greek Greek orthography, phonology, morphology, and epic hexameter

Kanōnes, a system for building Greek morphological parsers


  • Corpus specific: lexical stems and inflectional patterns can be easily tailored to analyze a specific corpus of texts
  • Citable: analyses identify the lexical entity and morphological analysis with URNs, and include URNs identifying the specific stem and inflectional rule used to arrive at an analysis
  • Requires no coding: build a parser specific to your corpus by editing tables of stems and inflectional patterns stored in simple .csv files
  • Cleanly separated from and designed for integration with other systems: structured output in JSON and RDF formats

Current status

In active development. Current emphasis is on developing large lexica of stems that users can draw from in building their own corpus-specific parsers. Tables of noun stems from the Perseus project’s morpheus parser have already been integrated into Kanōnes, and work is now focused on adding verb stems.

Documentation for end users

See a brief overview of using Kanōnes.

Release information