iStar modeling is beneficial in the early stage of requirements engineering, helping requirements analysts to analyze stakeholders’ intents and interrelationships. However, it is time-consuming to perform the iStar modeling manually, requiring (semi-)automated support. To facilitate the iStar modeling practices, we conduct ongoing research on a user-friendly semi-automatic iStar modeling approach, which aims to assist users in iStar modeling by extracting model elements from natural language requirement artifacts. Specifically, based on the analysis of the actual modeling process, this work proposes to automate three modeling steps: the actor entity extraction, the actor relation extraction, and the intention entity extraction. Then, this work proposes a hybrid approach for natural language processing to extract the model elements in requirement sentences to automate these three modeling steps. This hybrid approach consists of two parts: the deep learning-based method and the logical reasoning method, which utilizes both results simultaneously, ensuring the high accuracy and interpretability of the results. We preliminarily evaluated our proposed approach, and the results show that our proposed approach is efficient and helpful.