 This paper proposes a novel approach to diagnosing tinnitus using machine learning. The authors use meta-learning to improve the accuracy of diagnosis across multiple data sets, even when they have different characteristics such as age, gender and data collection methods. Additionally, the authors incorporate ear-side information into their model to enhance its performance. Their results demonstrate that this approach outperforms other existing methods with an accuracy of 73.8 percent. This article was authored by Zhilu, Yunli, Li Niao and others.