Considerable progress has been made in automating aspects of literature review, but current tools still struggle with complex tasks like efficient literature retrieval and topic analysis. Challenges such as ensuring relevance in retrieved studies and uncovering thematic relationships across large datasets remain inadequately addressed. To tackle these issues, we propose a framework that combines Research Rabbit with advanced NLP models, KeyBERT and BERTopic, to automate the retrieval and categorization of literature in systematic literature reviews (SLRs). We aim to streamline SLR studies’ time-consuming manual retrieval and categorization by using an embeddings-based comparison of keywords and researcher-provided search. Our framework operates in three stages: (1) retrieval of relevant literature using Research Rabbit, (2) keyword extraction with KeyBERT, and (3) thematic clustering via BERTopic. One distinctive feature of our framework is the comparison of embeddings from the generated keywords with those from the researcher-provided search strings. The framework uses cosine similarity to calculate a relevance percentage value, ensuring that only highly relevant studies are selected. We present a use case to demonstrate the application of the proposed framework. Our framework identified 49 key papers on a given topic, extracted thematic keywords, and provided insights into topic clusters. This automated process reduces manual effort, enhances search accuracy, and uncovers hidden research patterns, offering a scalable and efficient approach to SLRs.