Fig. 3

A KN representing knowledge triples found in the test set of 4023 PubMed articles. Yellow nodes refer to a trait entity, red nodes refer to gene entities, and green nodes represent the metabolite entities. a complete zoomed-out knowledge network b zoomed-in snapshot of the knowledge network focusing on tuber flesh color and additional traits with their respective biological associations. The color of edges reflects the document frequency (weight) of a relationship. Grey edges appear only in one document, whereas black edges appear in more than one. The nodes are organized in circles around traits of interest. Each of those traits has its 1st order neighbours in a circle around it (circle 1). The remaining nodes are organised with respect to the main trait of interest (flesh color nodes). The nodes in circles 2, 3 and 4 (counting from the center) are 2nd order neighbours of the flesh color node; the nodes in circles 5 and 6 are 3rd order neighbours of it; the nodes in circle 7, 8 (the outermost ones) are 4th (or higher) order neighbours of the flesh color node, or not connected to it at all