Skip to main content

Table 4 Results of self-training

From: University of Turku in the BioNLP'11 Shared Task

 

Random distribution (devel/test)

Even distribution (devel/test)

S 3

55.97%

56.17%

S 2

56.18%/52.72%

56.83%/53.21%

S 1

54.83%

55.78%

S 0

55.67%

55.79%

baseline

55.46%/52.84%

55.46%/52.84%

  1. Performance of the system on the GE subtask 1 in terms of F-score on the overall Approximate Span & Recursive matching criterion. Random distribution refers to self-training example selection by random sampling, whereas even distribution refers to selection of equal amount of examples for each event type and argument combination. Baseline is the performance of the system with no self-training (trained on GE subtask 1 data only).
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us