DISCOVERING IDEATIONS FOR CHATBOT THROUGH SOCIAL MEDIA MINING

• Main Author: Siwi Murwati, Anggun
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69810
• Institution: Institut Teknologi Bandung

Chatbot has been widely employed throughout company sizes and industries. Even though chatbot has proven to be far more efficient and quicker than human agents, it does not always give the most satisfying experience for customers since they lack personal touch. Customers are often left hanging and unsatisfied with chatbot services. To prevent customers from being disappointed with the service quality, firms need to understand what the customers need in order to improve chatbot’s performance. To achieve this goal, this study used a hybrid method of lexicon based TextBlob and Logistic Regression techniques to identify the sentiment of chatbot, followed by LDA Topic Modeling techniques to discover the most discussed topics from customers on Twitter and continued with ideations discovery. The result shows that 39% of customers’ sentiment towards chatbot is positive, while 34.9% express negative sentiments. This indicates a growing number of unhappy customers which demand serious attention. Furthermore, LDA topic modelling revealed eight topics that are most discussed by customers on Twitter such as Natural Dialogue, Convenience, and Service Quality. Having this set of topics as lead, this research contributes ideas which can be utilized by firms or developers to streamline the product innovation process of chatbot and and expected to provide an improved service quality to customers.