Sieve-based relation extraction of gene regulatory networks from biological literature

BMC Bioinformatics

Table 4 BioNLP 2013 GRN Shared Task development, training and test dataset properties.

Dataset	dev	train	test
Documents	48	86	67
Tokens	1321	2380	1874
Real mentions	205	422	290
Action mentions	55	102	86
Events	72	157	--
Relations	105	254	--
Interaction relations	71	159	--

The numbers of the Interaction relations that our system reads from the datasets is different than the real ones due to the import technique into our internal data representation. The dev dataset contains 67 and training dataset contains 131 reference Interaction relations. The test data contains 88 such relation instances (The number was retrieved from the output of the official BioNLP GRN Shared Task test evaluation service).

ISSN: 1471-2105