PetRobustness

Some notes on how to make parsing with Pet more robust.

TableOfContents

Unknown words

• POS -> generic

  • -default-les
    • PET mapping

• lexical type prediction

  • -predict-les
    • ME model, trained on treebank (2,000 + sentences) + list of lexical types pred-lex.tdl

• super tagger ?

• chart mapping/ Reg-ex token handling

Grammar Internal Solutions

• Roots (in english.set) Robustness rules/Mal rules

Pet settings

• always use packing
• recommend -memlimit (amount/2) rather than -limit (edges)
• -timeout=1 (second) can also be useful