Enhancing the Diversity of Smart Reply Suggestions: A Novel Approach Combining Text Classification and Post-Processing Techniques for Real Conversations in Bahasa Indonesia

Smart Reply is a Natural Language Processing application that offers suggestions for replies, enabling users to respond to messages quickly without having to type them out.However, the application sometimes generates reply suggestions that have similar meaning.In this research, we employ text clustering methods to address this problem.Furthermore, the application often provides specific information, such as phone numbers and addresses, which can be generated from the training data.To address this problem, we implement post-processing methods, including removing phone numbers or addresses from reply candidates and utilizing the most frequently used replies to introduce greater variety.Our results demonstrate the effectiveness of our approach in diversifying reply suggestions.