Integration of IndoLEM and IndoBERT Models in Realtime Question Prediction in Customer Service Support
Muhammad Haris, Al-Khowarizmi Al-Khowarizmi, Okvi Nugroho · 2024
This research has a research object, namely predicting questions for real-time support services for customers. This research applies the IndoLEM and IndoBERT integration models. The use of these two models can develop the scientific discipline of natural language processing (NLP) which focuses on Indonesian. In the IndoLEM process, it will be used to group questions based on labels and perform vector representation on the data, while IndoBERT functions to understand the context and provide more relevant predictions. Testing was carried out in two stages, with the results showing an increase in accuracy from 83% in the first test to 90% in the second test. This improvement shows that the integration of the two models is effective in improving the system's ability to answer customer questions accurately and quickly. Challenges such as data suitability and increasing system complexity can be addressed through comprehensive optimization and validation. The results of this research show that the combination of IndoLEM and IndoBERT can significantly improve the quality of customer service by providing more efficient and relevant solutions. Thus, this integration model has the potential to be widely implemented in customer service systems in various sectors. This success underscores the importance of utilizing advanced NLP technology to improve customer satisfaction and operational efficiency, making service systems more responsive and adaptive to user needs