Figure 6
From: Sieve-based relation extraction of gene regulatory networks from biological literature
Data flow in CRF-based relation extraction sieves. First, the initial skip-mention sequence is transformed into the selected skip-mention sequences. Then, for each of the skip-mention sequence type, a different CRF model is trained and then used to label the appropriate skip-mention sequences. After labeling, the relations are instantiated from the tagged sequences and returned as a result.