WPQA: a Lite Indonesian Question Answering for Wikipedia
Mohammad Yani, Cindy Apriliyani, Muhamad Mustamiin, and Rendi

Politeknik Negeri Indramayu

Abstract

In recent years, researchers in natural language processing have developed many pre-trained models based on BERT. One of these models is IndoBERT, an Indonesian-language version of BERT trained using Indonesian-language online articles. IndoBERT has indeed been used to evaluate Indonesian-language benchmark datasets. However, practical implementations are still limited, including the implementation of IndoBERT in a question-answering system application to obtain information from Wikipedia. In this study, the author proposes a simple Indonesian-language question-answering system called the Lite Indonesian Questions Answering for Wikipedia (WPQA), which can generate answers from Wikipedia for a given question. This system consists of three components: input, process, and output. This system inputs keywords and questions. Then, the input is processed using the IndoBERT model to search for answers from the given context. The result is an answer obtained from the context search. The system is evaluated by using a psychometric tool to measure users^ perceptions of the system. In the evaluation, we have a user satisfaction rate of 82%, who state that they are satisfied with the system. The percentage is measured by using Likert scale. The evaluation results show that WPQA can be used to search for information on Wikipedia topics from a Wikipedia context.

Keywords: question answering, wikipedia indonesia, indobert

Topic: Artificial Intelligence (AI)