Evaluating Turkish BERT-based Language Models for Effective Customer Feedback Interpretation in CRM

Customer relationship management (CRM) solutions are crucial in modern business strategies for managing and strengthening customer connections, increasing customer satisfaction, and generating long-term corporate growth. By combining advanced machine learning algorithms with CRM platforms, this paper investigates the possibility of generating more precise customer satisfaction ratings. By utilizing deep learning and natural language processing (NLP) techniques, organizations can effectively evaluate customer feedback and boost client satisfaction. This study demonstrates the efficacy of NLP-powered sentiment analysis in interpreting customer complaints and promptly addressing customer satisfaction issues. Entity identification is also beneficial for identifying patterns in client feedback, which can be utilized to enhance the product and attract new customers. The findings of our study highlight the superior performance of the ElecTRa model over other evaluated models in terms of accuracy, precision, recall, and Fl-score; specifically, th e El ecTRa model ex hibited an average ROC AUC score of 0.93. Automatically classifying customer notifications as complaints or non-complaints enables the monitoring of customer representatives' performance, facilitating targeted training in-terventions to address any identified shortcomings and improve service quality. By pinpointing specific issues where mistakes were made, organizations can effectively tailor training programs to enhance the skills and effectiveness of customer representatives in handling customer concerns.