The Comparison of Semantic Suffix Tree Clustering and Suffix Tree Clustering Algorithm Influence on the Accuracy Rate of an Indonesian Question Answering System

摘要

This research analyses the comparison between Semantic Suffix Tree Clustering (SSTC) algorithm and Suffix Tree Clustering (STC) algorithm in clustering documents on Indonesian Question Answering System. The fundamental difference between the two algorithms is that SSTC considers the meaning of the words to generate clusters and readable labels rather than relying only on string matching like STC. As such, SSTC is able to return a more specific and accurate clusters, and has potential to increase the accuracy rate of a question answering system. However, contrary to the hypothesis, comparison results shows that the accuracy rate of Indonesian Question Answering System after the documents is clustered by SSTC is lower than by STC. The accuracy rate degradation occurred in almost every question category, except Definition category. In average, the accuracy rate obtained by Indonesian Question Answering System with SSTC is only 23.31%, while Indonesian Question Answering System with STC is able to obtain 83% accuracy rate. This significant difference indicates that Semantic Suffix Tree Clustering algorithm is not suitable in the context of document clustering on Indonesian Question Answering System.