On Nov 1, 2012, at 10:57 PM, Tim Miller <[email protected]> wrote: > So I've tracked down the issue where POS tags were null in the > dependency parser, but not when the Chunker was added to the pipeline. > The issue was that the POSTagger uses the subiterator method, which > requires priorityList section in the descriptor to work. The POSTagger > descriptor is missing it but the Chunker descriptor has it, so loading > the Chunker descriptor was allowing the subiterator call in POSTagger to > work. > > With me so far? I will check in the change to the POS descriptor. But > to fully fix I had to re-generate the dependency parser descriptors > (using the WriteClearParserDescriptors class). This results in some > additional changes to those which are a result of the type priority > change (see diff below). Is it safe to check in all of that or just the > POS descriptor change and expect others to generate their own dependency > descriptors?
Do we have a test that runs the dependency parser from the descriptor? If we do, and the test passes after these changes, I'd say go ahead. > - <import > name="org.apache.ctakes.typesystem.types.TypeSystem"/> > + <import > location="../../../ctakes-type-system/src/main/resources/org/apache/ctakes/typesystem/types/TypeSystem.xml"/> This is the only part that looks suspicious. We want by-name imports, not by-location imports, in general. Steve
