Abstract
Abstract
Legal professionals often face challenges in efficiently accessing relevant information from extensive and complex legal documents. This thesis presents a method for passage retrieval in Turkish legal texts utilizing the Bidirectional Encoder Representations from Transformers (BERT) model. The research aims to enhance the retrieval process by leveraging contextual embeddings to understand the nuanced language and terminologies used in legal documents. The study involves the creation of a dataset from Turkish legal books and the application of both BM25 and BERT models for retrieval tasks. Results indicate that the combined use of BM25 and BERT improves the accuracy and relevance of retrieved passages, offering a promising tool for legal research in the Turkish context.