Due to the swift progress in computing and information technologies, traditional access control frameworks struggle to encompass the subtle security needs of emerging applications. An alternative, the attribute-based access control (ABAC) model, offers greater adaptability in meeting the authorization demands of intricate and evolving systems. This paper presents a framework suggesting the application of advanced Natural Language Processing (NLP) techniques to generate embedding vectors for attribute requests, employing a Skip-gram architecture to capture contextual nuances. With well-defined features extracted from these requests, a robustly trained machine learning classifier authorizes each request appropriately. We envision that embedding layers could enhance and potentially become integral to traditional access control models, aiding in capturing nuanced features over time. Through experimentation on real-world datasets, we validate the feasibility of our approach, benchmarking it against the State-of-the-Art (SOTA) model. Additionally, we discuss the challenges encountered and propose avenues for future research.