A corpus-based sub-graph
First, a separate sub-graph for each primary stimulus was created for each text in the corpus. All sub-graphs were obtained with empirically adjusted parameters [HAR 14], such as: intermediate nodes in the path of l = 3 for the extracting algorithm, and direct associations with the stimulus minimum dAn = 2, with a sub-graph adjusting parameter of Ss = 0.5 for the control procedure. Then, the text-based sub-graphs obtained for a specific primary stimulus were merged into the corpus-based primary stimulus sub-graph, i.e. all sets of nodes and all sets of edges were merged, forming a multiple set union. Finally, the corpus-based primary stimulus sub-graph was trimmed, which means that each non-connected node was removed from the final subgraph, and each open path (paths with which the end node had not connected) which had more than two edges between the stimulus and the end node was reduced to match the network-forming principle that a stimulus (A) produces an association (B), which then serves as a stimulus to produce an association (C). Afterwards the reduced path takes the form A - B - C.