Kuzushiji OCR using machine learning (deep learning) has attracted many people’s interests. To push forward this line of research, we held a machine learning competition “kuzushiji recognition” between July and October, 2019 on Kaggle, which is globally the largest platform for machine learning competitions. This paper summarizes what we did and what we learn from the competition on the preparation, progress, and results stages, and discusses the value of machine learning competitions in digital humanities research.

      @InProceedings{ jm19a,
      author = {Asanobu KITAMOTO and Tarin CLANUWAT and Alex LAMB and Mikel BOBER-IRIZAR},
      title = {Progress and Results of Kaggle Machine Learning Competition for Kuzushiji Recognition},
      booktitle = {Proceedings of IPSJ SIG Computers and the Humanities Symposium 2019},
      pages = {223-230},
      year = 2019,
      month = 12,
      doi = {},
      note = { (in Japanese)},
      }