Fig. 2From: Relation extraction between bacteria and biotopes from biomedical texts with attention mechanisms and domain-specific contextual representationsBox-and-whisker plot of average F1 score distributions of the deep-learning-based relation extraction models on the BB task. The comparison between our model and existing deep-learning-based models is shown in blue; the improvement of our model after adding each of the proposed contributions is shown in grey. Note: “Attns” denotes the use of integrated attention mechanismsBack to article page