In this paper, we conduct research on Vietnamese spelling correction. There has been a lot of research on this topic either using statistical language models or deep learning models. With statistical language models, one of its biggest challenges is multiple errors in a row, which negatively affects how models can detect errors. With deep learning, many state-of-the-art methods have been proposed using sequence-to-sequence models. Despite the improvement, we found that these models still have a lot of difficulty in dealing with out of vocab words, which results in false detection and correction. Therefore, we proposed a solution to enhance the model’s ability to deal with these cases using Pointer Net and Transformer. Our model has achieved an encouraging result compared to other state-of-the-art models. Our implementation is available at: https://github.com/hinetabi/Pointer-Network .