Pattern Combination and Variation Syntax for Case Similarity

Abstract

This paper aims to explain a new way of seeking case similarities to Case Based Reasoning question and answering system. A new concept used is to create a combination of sentence patterns with the variations of syntax and relation between words in sentences. More precisely, we improve the similarity from a resource-poor source language. The combination of pattern was made for the variation of the fridge using Predicate Argument Structure (PAS) analysis, whereas the combination of pattern with the relation between words was by forming Sub Tree (ST) and Subset Tree (SST) from parsing tree. The result of forming a combination of these patterns was then selected to obtain an appropriate answer. Accuracy resulted by using the combination of this pattern resulted in a significant value equal to 88.64% of 132 question sentences used as data testing.