Augmenting Legal Reasoning with BERT: The Second Iteration of Bekenbey AI

In recent years, the rapid advancement of technology has necessitated the use of machine learning models and artificial intelligence systems not only in the field of engineering but also in the resolution of complex problems in social sciences. In this study, we build upon the Bekenbey AI model, which is the first study in the literature to establish a connection between the legal domain and Generative AI. While the foundational model showed significant accuracy and robustness across various legal applications, recent advancements in natural language processing (NLP)—particularly in Transformer-based approaches—present new opportunities to enhance the model’s performance and adaptability. Our primary objective in this version is to integrate Bidirectional Encoder Representations from Transformers (BERT) into the existing Bekenbey AI pipeline to refine legal text processing, improve document classification, and provide more nuanced predictive analytics. By benefiting from BERT’s self-attention mechanisms and its powerful contextual embedding capabilities, this study aims to handle the complex semantic structures inherent in legal language with greater precision. The proposed approach enhances both classification accuracy and text generation quality, demonstrating the effectiveness of integrating domain-specific attention mechanisms with transformer-based models in legal AI applications. Experimental results indicate that the proposed model consistently outperforms the baseline across multiple evaluation metrics.