A FRAMEWORK OF QUESTION ANSWERING SYSTEMS FOR DIABETES CARE USING LATENT SEMANTIC INDEXING WITH TEXT MINING

Abstract

Currently, question answering systems still have some problems due to the ambiguity of words. Sometimes, the words with the same meaning, but differently writing can bring the wrong answers. Latent Semantic Indexing (LSI) is one method that many researchers used to solve a problem of synonym since LSI can be applied for finding the latent semantic of the synonym. Moreover, LSI also reduces the document size while their meaning remains. This paper presents a conceptual framework for the development of a question answering system using LSI. Here we applied the question answering system for diabetes care.  The framework consists of three main steps, i.e., (1) document pre-processing, which is applied by a technique of text mining, (2) LSI answer scoring, which follows LSI methods by term frequency-inverse document frequency (TF-IDF) weighting, and (3) question answering matching, which use the similarity measurement. This paper also includes examples of each step. A preliminary experiment shows that the conceptual framework offered can provide the correct answer.