IDEA #1HKYOL Topic Transfer Technique. 23A0034

Key Words: Natural Language Processing, Contextualized Topic Modeling, Large Language Models, Transfer Learning The purpose of our invention is to build models to derive topics from known textual sources in order to apply those models to unknown and unclassified text, to derive the topics used in that text... To the best of our knowledge, this “topic transfer” from one data set to another (which we used in our study from legitimate forums to underground forums) is a unique feature of our approach even in the NLP context. The notion of topic transfer introduced in our work should not be confused with transfer learning, a much broader set of AI techniques used, e.g., in the topic modeling step of our method.