Khoo manually constructed a set of graphical patterns that indicate the presence of a causal relation in sentences, and which part of sentence represents the cause or the effect (Khoo, 2000). These patterns are matched with the syntactic parse trees of sentences, and the parts of the parse tree that match with the slots in the patterns are extracted as the cause or the effect. Khoo applied 68 graphical patterns on 100 medical abstracts in the Medline database [MEDINE 2001], and reported the unsatisfactory result with an accuracy of around 50%. This low-precision problem requires an additional component for pruning extracted relations. A popular way is to incorporate a classifier trained with supervision (Zhang, 2014).