Scholarships have played an important role in improving education by offering opportunities for students pursuing academic achievement, financially supporting talented students from diverse backgrounds, and lifting financial burdens. However, the traditional process of distributing the scholarships has many challenges. These manual processes lack transparency, immutability, and audibility, leading to inefficiencies such as being prone to mistakes, being time-consuming, and losing students’ documents. In this study, we introduce blockchain and sentiment analysis to solve those issues and heterogeneous data limitations. Specifically, we utilize IBFT 2.0 proof of authority blockchain consensus and a pre-trained transformer learning model with sentiment analysis to enhance the scholarship approval manual process. We employ pre-trained DistilRoBERTa transformer learning with a categorized sentiment analysis model to address the heterogeneous data limitation. We evaluate the students’ unstructured data and classify it as positive, negative, or neutral. Moreover, we adopt the advanced encryption standard (AES) with interplanetary file systems (IPFS) to protect the student’s sensitive information. The proposed method enhances the centralized system by making it faster, more auditable, and more accurate. Specifically, our experimental results show that our approach could achieve a throughput of 0.25 transactions per second for 10 and 20 students and 0.24 transactions per second for 50 and 100 students. Additionally, the time consumption results indicate that the model takes 242.30, 482.69, 1208.07, and 2419.56 s to process 10, 20, 50, and 100 students, respectively.