As technology continues to advance, the necessity of storing information in digital environments has become increasingly evident, leading to a substantial growth in data size within these systems. Quick access to information and efficient learning have made digital resources the preferred choice for most researchers. However, reading and analyzing lengthy scientific publications remains a time-consuming task. In recent years, automatic summarization systems have provided valuable support to researchers in this area. Among these systems, abstraction-based methods, which mimic human analysis by interpreting original texts and rephrasing them with new words, are particularly challenging yet effective. This study focused on automatically summarizing scientific publications using a deep learning-based approach. The generated summaries were evaluated against human-written summaries on criteria such as coherence and length. The unique challenges of the Turkish language, compared to other languages, add further importance to this work. The evaluation yielded promising results, with precision scores as high as 0.93 and F1-scores reaching 0.57. Moreover, the ability of the system to produce summaries closely resembling those created by humans highlights the effectiveness of the proposed methodology. These findings hold significant potential for advancing the development of automatic summarization systems tailored to the Turkish language.