Prediction of Turkish Constitutional Court Decisions in Terms of Admissibility and Violation of Rights With Artificial Intelligence

The digitization of legal texts and advances in information processing theories and technologies have triggered several transformations in both the practice and teaching of law in recent years. Techniques developed in areas such as artificial intelligence, natural language processing, text mining, and machine learning have drawn the attention of legal practitioners and academics to this field. It is possible to remove obstacles to access to justice, improve legal security and certainty, and solve practical problems faced by legal practitioners by employing assistive tools to be created by using artificial intelligence technologies in the field of law. This study aims to develop an algorithm to predict the results of the Constitutional Court of the Republic of Türkiye on individual applications in terms of admissibility and whether there is a violation of rights by using machine learning and natural language processing techniques. In the study, the texts in the "Facts" title of the reference texts were used. A success rate of 91.56% was achieved for admissibility and 97.18% for whether there was a violation of rights. The study is unique in its field in that it performs a two-stage prediction task regarding admissibility and merit, provides a highly representative model since it includes all processable data, does not use a data augmentation method, and has a high success rate.