Skip to main content
Fig. 3 | BMC Bioinformatics

Fig. 3

From: BioReader: a text mining tool for performing classification of biomedical literature

Fig. 3

Results pertaining to classification of articles relating to infectious diseases vs. non-infectious diseases (allergy, autoimmunity, cancer, etc.) using a glmnet classifier. a) BioReader learning curve for five-fold cross-validation with glmnet on corpora ranging from 50 to 1500 abstracts in intervals of 10 abstracts (average over 100 iterations). b) ROC curves of performance of BioReader and MedlineRanker trained with 1500 abstracts and evaluated on 500 abstracts excluded from the training. c) BioReader F1 scores for positive and negative abstract classification at varying proportions of training set size (total 750 abstracts) for each category in intervals of 10 abstracts (average over 100 iterations). The classifier was applied to a balanced test set of 500 abstracts

Back to article page
\