Russian Court Sentences Texts Summarization using Abstractive Summarization Techniques

Master thesis topic: Russian Court Sentences Texts Summarization using Abstractive Summarization Techniques Although court sentences texts have a clear structure, like many other legal documents, they are difficult to read, contain a lot of additional information and complex syntactic structures, therefore, difficulties arise when the reader just wants to get only necessary information and understand what the essence of the crime is. This study observes the works of authors who had previously conducted research in the field of text summarization (both classical methods and methods using neural networks were considered), a dataset containing Russian-language texts of court sentences and their summaries was created, an overview of the architecture of the models used was presented, the analysis of the texts of court sentences, the details of preprocessing, the selected parameters and metrics, and the results obtained were described. The best results were performed by MT5 model.