GeneTag (1) is a tagged corpus for gene/protein named entity recognition.

The corpus is annotated as follows:

The_TAG concentration_TAG of_TAG alpha_GENE1 2_GENE1 -_GENE1 macroglobulin_GENE1 ,_TAG alpha_GENE2 1_GENE2 -_GENE2 antitrypsin_GENE2 ,_TAG plasminogen_GENE1 ,_TAG C3_GENE2 -_GENE2 complement_GENE2 ,_TAG fibrinogen_GENE1 degradation_GENE1 products_GENE1 (_TAG FDP_GENE2 )_TAG and_TAG fibrinolytic_TAG activity_TAG [...]

The corpus is stored as a list of instances each consisting of (a) an identifier on the first line, and (b) a sequence of tokens on the second line. Each token is of the form word tag where tag can be one of {TAG, GENE1, GENE2}. TAG represents any word that is not part of a gene name, while GENE1 and GENE2 mean the token is part of a gene name. The index 1 and 2 just works as indicators to distinguish two dierent but adjacent gene names.

Why not using the BIO notation (or something similar)? E.g.:

The_O concentration_O of_O alpha_B 2_I -_I macroglobulin_I ,_O alpha_B 1_I -_I antitrypsin_I ,_O plasminogen_I ,_O C3_B -_I complement_I ,_O fibrinogen_B degradation_I products_I (_O FDP_B )_O and_O fibrinolytic_O activity_O [...]

(1) Tanabe, Lorraine, et al. "GENETAG: a tagged corpus for gene/protein named entity recognition." BMC bioinformatics 6.Suppl 1 (2005): S3.

  • But why should it? – curiousdannii Oct 15 '15 at 5:54
  • @curiousdannii if you train a tagger on it, say an HMM or CRF, you'll have two classes gene1 and gene2, while it would make more sense to have a class begin and a class inside. Also BIO is pretty much the standard de facto for NER annotations. – Franck Dernoncourt Oct 15 '15 at 6:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.