@article{article_1666175, title={Artificial Intelligence-Based Automation of the Referral Process for Applications Submitted to CİMER}, journal={İletişim ve Diplomasi}, pages={175–200}, year={2025}, DOI={10.54722/iletisimvediplomasi.1666175}, author={Özalp, Abdulkadir}, keywords={Cumhurbaşkanlığı İletişim Merkezi, yönetişim, otomasyon, kamu kaynaklarından tasarruf, yapay zekâ}, abstract={In recent years, technological advancements have significantly increased the volume of data stored and processed. While this growth presents many advantages, it also brings challenges, such as the need for effective text classification. In Türkiye, the Presidency’s Communication Centre (CİMER) was established to promote principles of good governance, such as accountability, transparency, the rule of law, and citizen participation. CİMER is an effective channel through which citizens can obtain redress for administrative actions, and the number of applications submitted has been increasing annually. As the number of citizen applications submitted to CİMER increases each year, addressing each application within the legally mandated timeframe has become increasingly demanding. In this context, handling all procedures related to the referral of CİMER applications within an automated system is very important. In addition, the manual referral of applications to the relevant public institutions places a considerable burden on human resources. This study introduces a novel approach using artificial intelligence to automate the referral process of CİMER applications. It proposes a system in which applications submitted to CİMER are classified by a pre-trained artificial intelligence model operating in the background of the CİMER system. Based on the classification results, applications are either automatically forwarded to the relevant ministry or sent to the CİMER application pool for manual referral. The study compares two deep learning methods for text classification—Convolutional Neural Networks (CNN) and BERT. The analyses show that the BERT model outperforms CNN, achieving a validation accuracy of 99.986% and a test accuracy of 99.924%.}, number={14}, publisher={Cumhurbaşkanlığı İletişim Başkanlığı}