List view
Improvements to POS component. Determine that POS models are trained correctly, using newer tokenization. Train using old corpora and new corpora, evaluate using tokenization with and without hyphen splitting, and confirm performance is as expected on new corpus when hyphen splitting is active. Update POS, pipeline config. Deploy new POS models. Additionally, any very minor fixes to POS code.
Overdue by 7 year(s)•Due by June 15, 2018•0/2 issues closedThis milestone is for tracking issues that are planned for the next Major version release. Especially changes that break backwards compatibility.
No due date•3/3 issues closed