PACIS 2019 Proceedings


Transfer learning has been widely utilized in many real-world applications, including text analytics. However, our literature review suggests that it could be beneficial to conduct a more comprehensive study on transfer learning and its application. Grounded by theory of transfer of learning, we identify and re-organize different types of transfer learning approaches in a more systematical and theoretical manner. We believe different transfer learning approaches are complementary for one another rather than substitutive. As deep learning has achieved great successes and been applied in solving various healthcare related tasks, we propose a design framework that integrates deep learning model and three different transfer learning approaches and evaluate the performance of our framework in a healthcare text analytics task. Our preliminary results confirm the improved performance by incorporating the three transfer learning approaches individually and in combined. We also discuss the theoretical and practical implications and research plans.