Mahidol University has made it a priority to conduct high-quality research to meet the needs of a rapidly changing world and respond to both Thailand’s and the world’s problems. Artificial Intelligence (AI) technology has become more prevalent in everyday life than ever. The AI technology is often used to facilitate the discovery of knowledge in large databases such as “Wikipedia”. The AI technology can assist in automatically retrieving answers to natural-language questions. With such capability, users no longer have to waste time searching for the answers on their own.
WabiQA is a novel system for automatically answering questions in the Thai language utilizing the Thai Wikipedia articles as the knowledge source. It was conducted by Miss Lalita Lowphansirikul, Miss Yodtida Yodmuang, and Miss Wannakan Prangaon, under supervision of Assoc. Prof. Dr. Suppawong Tuarob, Head of the Machine Intelligence and Knowledge Engineering (MIKE Cluster), and Dr. Thanapon Noraset, instructor.
WabiQA is a novel system for automatically answering questions in the Thai language utilizing the Thai Wikipedia articles as the knowledge source. WabiQA takes questions in the natural Thai language, such as “When was the Faculty of ICT, Mahidol University established?” as input, then displays the answer, e.g., “20 May 2009,” along with the specific document in which the answer is found. However, all answers must be in the (Thai language) Wikipedia database only (https://th.wikipedia.org/wiki).
WabiQA implements the BM25F-based retriever to identify articles on Wikipedia that are most likely to contain the answer or are related to a given question. Then, a Bi-Directional Long-Short Term Memory (BiLSTM) model is applied to read the Thai article and locate candidate answers. Lastly, the Attention layers are used as general answer predictors to quantify the confidence that the target text is the answer to the input question. The test result showed that the WabiQA was able to reduce search time by 97.81 percent. Furthermore, the research team also developed a prototype mobile application that aims to facilitate Thai users with visual impairments using voice-to-speech technology and an intelligent question-answer categorization.
This work was published in the Information Processing & Management Journal, a leading international academic journal ranked Quartile 1 (Q1) in Information Systems. For those who are interested in the full article, please visit: https://doi.org/10.1016/j.ipm.2020.102431.
Dr. Thanapon Noraset, said, “This research is a collaboration among the Faculty of ICT’s research team, students, and partners, that aims to develop artificial intelligence research and innovations, driven by the technological capabilities of the Thai people, to address difficult real-world problems. Personally, I am impressed by the ability of Mahidol students to conduct such world-class research. In this study, we applied advanced artificial intelligence technology to learn and understand the Thai context. Furthermore, this technology can also be used to perform an automated analysis of large and heterogeneous data composed in Thai, such as searching and summarizing social media opinions on products, events, or policies.”
Did you know? WabiQA won the first prize award from Thailand’s 21st National Software Contest 2019 (NSC) under the category “Question-Answering Program from Thai Wikipedia.”
