Building a question classification model for a malay question answering system

Abstract

Question answering system (QAS) is an example of an application of natural language processing where it is able to automatically return a specific answer to a question given in a natural language by a human. One of the important tasks in QAS is Question Classification which is the task to identify the semantic type of the required answer for the question posed to the QAS. Identifying the correct answer type is an important process before the required correct answer can be retrieved by the system. In this paper we presents a model of Answer Type Classification using machine learning approach targeted for a Malay QAS for the Quran, which is a restricted-domain QAS. The performance of the classification model using three different machine learning classification algorithms, namely Naïve Bayes, Random Forest and Support Vector Machine (SVM), are then evaluated. The results show that the classifier based on SVM has the best overall results in terms of accuracy, precision, recall and F1-score.