Artifact – author matching in Turkish texts with stylometry analysis using natural language processing and machine learning methods

With the widespread use of technology and the Internet, changes and developments have occurred in many areas, including the field of journalism. Online newspapers have attracted the attention of many newspaper owners and writers who serve in this field due to their ability to incorporate various elements and appeal to larger audiences. In this context, traditional newspapers that were operating with conventional methods quickly became part of this process of change in order to provide services in the digital environment, resulting in a significant increase in the number of media organizations operating on the Internet. These developments in journalism have enabled writers working in different fields to reach readers on the same platform, leading to numerous works being produced in various areas. However, alongside these positive developments, the increase in data volume has made it difficult to access the desired and accurate data. As a result, extracting meaningful data from meaningless data and classifying data according to specific attributes have become important topics. In this study, articles written by individuals who engage in newspaper writing in the electronic environment were analyzed using methods of numerical style analysis, natural language processing (a subfield of artificial intelligence), and machine learning to accurately match the authors of these articles. Data preprocessing was performed on the dataset to optimize it, followed by feature extraction using the functions of the Zemberek library for natural language processing. Finally, a comparison was made among the machine learning classification algorithms used in the study to determine which one was more successful in author-work matching, and the model selection was intended to be determined accordingly.