Latin resource for CSTlemma
---------------------------

The flexrule file lemmatises all training words correctly and about 44% of OOV words wrong. (The training set, merely 18 thousand words/lemma pairs, is clearly way too small!)

Use the file called 'flexrules' together with the CSTlemma program. See https://github.com/kuhumcst/cstlemma

The flex rules were created on the basis of http://nlp.perseus.tufts.edu/syntax/treebank/ldt/1.5/ldt-1.5.tar.gz using the software found here: https://github.com/kuhumcst/affixtrain

Bart Jongejan 2015.11.12