Skip to main content

Table 10 Reasons for recall errors

From: Benchmarking natural-language parsers for biological applications using dependency graphs

Parser 1 bad tag 2 bad tags 1 missing 2 missing
Charniak-Lease 20.6% 2.0% 0.4% 0.0%
Bikel (0.9.8) 28.6% 3.4% 0.4% 0.0%
  1. For each dependency from the gold standard that was not generated by each parser, we determined whether one or both of the words it should have joined were badly POS-tagged, or entirely missing from the dependency graph of the parser's output.