This work is joint research with Suji Lee, Seokjin Han, Sewon Park, and Jaeyong Lee, 2017 - 2019.
In this project, we proposed an end-to-end deep learning model combining BNN with Korean speech recognition. Specifically, we combined BNN to implement the end-to-end model and obtain Monte Carlo estimates. We carried out experiments on the online dictionary dataset. In this project, I studied algorithms for inference in BNN models. We implemented Variational Dropout (Gal, Y., & Ghahramani, Z., 2016) to develop a model that can infer uncertainty in Korean natural language recognition problems. We presented the results of this project at Eastern Asia Chapter of the International Society for Bayesian Analysis and published in The Korean Journal of Applied Statistics in 2018.
Publication
S. Lee, S. Han, S. Park, K. Lee, and J. Lee. (2019). Korean speech recognition using deep learning. The Korean Journal of Applied Statistics, 32(2), 213-227.