An Effective and Interactive Training Data Collection Method for Early-Modern Japanese Printed Character Recognition

摘要

In this paper, we present a web application that supports to collect training data efficiently for early-modern Japanese printed character recognition. The national diet library in Japan provides a lot of early-modern (AD1868-1945) Japanese printed books to the public, but full-text search is essentially impossible. In order to perform advanced search in historical literatures, it is required extracting texts from images. To solve this problem, we have already proposed a multi-font Kanji character recognition method using the PDC feature and an SVM. For growing inperformance of this method, we need big amounts of training data. However, collecting training data by hand is extremely inefficient. Therefore, we propose a Web application that supports collecting training data.