The rapid increase in scientific literature has made it more challenging to effectively profile researchers based on their academic outputs. This paper explores and compares three keyword extraction techniques–TF-IDF, CF-IDF, and SBERT+KeyBERT–to enhance author profiling in academic environments. Using real data from Scopus via the Elsevier API, we measure the effectiveness of each technique using precision, recall, coincidence ratio, and cosine similarity (for SBERT+KeyBERT) to assess how well the models capture and represent an author’s research interests. Results demonstrate that while TF-IDF and CF-IDF offer solid keyword extraction based on term frequency, they struggle with semantic context. In contrast, SBERT+KeyBERT provides a more nuanced understanding of academic content by leveraging semantic embeddings, outperforming the traditional methods in terms of contextual relevance and keyword accuracy. These findings suggest that integrating SBERT+KeyBERT into academic profiling and recommender systems can significantly improve personalization and information retrieval in scholarly databases.