Designing answer sets for Thai advisory chatbot with different talking styles for COVID-19 pandemic

Abstract

This research focuses on designing a set of answers for Thai advisory chatbot using a different conversational approach from conventional chatbot for COVID-19 pandemic in order to improve user’s chatbot experience. The research studies how to modify the underlying interactive chatbot language to suit those extrovert and introvert personalities. To evaluate the correctness of answer set design, the author builds two classification models based on Extroversion Text Classification Model (ETCM). Both models utilize the same Logistic Regression algorithm, but different feature selection techniques. The first model relies solely on Term Frequency-Inverse Document Frequency (TF-IDF) property. In contrast, the second model utilizes both TF-IDF and a unique word component score for each speech segment as the principal features. Performance comparison of the study shows that the model utilizing both features is more accurate than that uses only TF-IDF.