Extraction of Predicate-Argument Structure From Sentence Based on PICO Frames

Identifying a logical relation between sentences and semantic role labelling requires a deeper knowledge of recognizing the relationship of various expressions. One method that can be used is by means of the extraction of predicate argument structure. This paper is purposely to describe a new automatic method for the extraction of Indonesian medical predicate-argument (P-A) structure analysis based upon PICO frame. Learning some relevant features, the method assigns some case roles (such as Problem/Population/Patient, Intervention, Compare/Control and Outcome) to the argument of the target predicate using the features of the words that are located closest to the target predicate. In this paper the illustration of their use in a pattern-based relation extraction component of PICO frame has been described. It is indicated from the test results that the use of the features with more semantic role categories in determining the P-A structure represents the respective results reaching at 89.35% for precision, 89.12% for recall and 89.98% for F1.