Masaki Nakagawa

Date of birth: Oct. 31, 1954. Place of birth: Tokushima, Japan.
Current position: Chief University Research Administrator of the University Research Administration Center,
Appointed Professor, Emeritus Professor, Tokyo Univ. of Agri. & Tech (TUAT).

Research, Education and Social Contributions

Professor Nakagawa has been working on handwriting recognition, pen/touch-based user interfaces and their educational applications.

In the 1990s, he actively collaborated with many companies. He led a project to create a large database of on-line handwritten Japanese character patterns with 16 companies and 7 university laboratories. This resulted in the Kuchibue database collected from120 people contributing 12,000 patterns each, which were collected mostly in their sentential contexts, and verified by both machine and human inspections. A portion of this database was freely made available for research. Subsequently, He made another database Nakayosi with 163 participants contributing 10,000 patterns each. So far, more than 70 groups have subscribed to them including 30 groups from abroad.

He collaborated with Fujitsu to combine on-line and off-line recognizers. Off-line recognizers are insensitive to wrong stroke order and overwriting, whereas on-line recognizers perform well for cursive or deformed patterns as long as their stroke orders are correct. These two approaches complement each other. His group was the first to implement a hybrid system for on-line Japanese handwriting recognition with a significantly improved recognition rate. They achieved the highest recognition rate for the standard database of Kuchibue. Even when the off-line prototype dictionary was compressed from 91.8 MB to 9.7 MB, their system achieved almost the same performance: 98.6% correct recognition rate when combined with contextual post-processing. In addition, he collaborated with Hitachi to create a Front-End Processor (FEP) for inputting handwritten Japanese characters for the most popular Japanese word processor "Ichitaro"

Pen interfaces are suitable for creative work that is not affected by input methods. He proposed Lazy Recognition (recognition results are not shown until needed) in the early 90s. He then presented prototype applications for PDAs, desktop tablet PCs, and large electronic whiteboards. He proposed a GUI for the large whiteboards that allows the user to manipulate the contents with an electronic marker from any standing position and attract his/her audience's attention without hiding the board by his/her body. He also presented several educational applications for the large whiteboards. This research received an award from the Japanese Society for Engineering Education in 2001. He collaborated with Hitachi Soft, which is now selling these products worldwide. He has received U.S. patents to scroll the window in proportion to the pen’s speed known as touch-scroll(U.S. Patent No. 6,128,014,“Human Interactive Type Display System,” filed on July 10, 1997, and patented on Oct. 3, 2000, U.S. Patent No. 6,683,628,B1,“Human Interactive Type Display System,” filed on May 11, 2000, and patented on Jan. 27, 2004, and others). These patents were initially owned by the university, but were later sold to a company for a significant price (highest amount among all the Japanese universities in the fiscal year 2010).

From around 2000 to 2010, he worked on removing writing constraints as much as possible. Until then, Japanese text had to be written neatly with one character within each writing box to remove segmentation problem, because it is written without space between characters, and some characters are composed of multiple radicals that are themselves characters. He proposed a recognition model, methods and practical systems. The model evaluates the likelihood based on character segmentation, character recognition, character pattern structure and context. A recognition system based on this model separates freely written text into text line elements, estimates the average character size of each element, hypothetically segments it into characters using geometric features, applies character recognition to segmented patterns, and employs the model to search the text interpretation that maximizes likelihood as Japanese text. The writing-box-free recognizer based on the model was used for several products through a university-based start-up, iLabo Corporation. Samsung's GALAXY for NTT docomo's smart phones, NEC tablets, and SONY tablets are some of the devices that incorporated the recognizer as a standard input method.

For the past 20 years, he has been working on the computer processing of historical documents. He applied the handwriting recognition technology to a dictionary of ancient cursive scripts of 67,739 categories and made them searchable by writing a pattern with a pen on a tablet or with a mouse, which received an award by a Japanese engineering newspaper (Nikkan Kogyo). He has been also working on image processing and character recognition to help archaeologists to read wooden tablets excavated from the Heijo Palace Site of 1400 years ago in Nara. This has a large cultural value but private sectors are reluctant to join because of smaller budgets compared with the OCR business. He has expanded this approach to historical document processing to Vietnamese old documents in Chunom, and to medieval and early modern handwritten and wood-printed text in Japan.

In the recent ten years, he has been working on applying Deep Neural Networks for recognizing both online and offline handwritten patterns of Japanese, Chinese, and Vietnamese, and of mathematical expressions. CNNs and RNNs have surpassed the performance of previous handmade recognizers on many problems with simpler structures. He used previously created databases for this research, and also published a large online database of handwritten Vietnamese characters. The research results have been published actively in top journals and conferences for pattern recognition and many of them are being transformed into products and services by iLabo and others.

He designed a new M.Sc. course on Ubiquitous and Universal IT Environment Design supported by the Japanese Ministry of Education (MEXT), which ran from 2005 to 2011; and an Education Program of IT Engineers for Advanced Manufacturing supported by MEXT and METI (Ministry of Economy, Trade and Industry), which ran from 2008 to 2014. A graduate of the first course established a start-up company Yappli, which is now listed in the Tokyo stock market. As a part of these programs, he set up exchange agreement with top engineering schools in Vietnam, Thailand, China and so on. These courses also enabled him to supervise 20 PhD graduates from abroad among a total of 40 PhD graduates. After funding for these programs ran out, they are succeeded in the current IT department of TUAT.

He has been promoting collaborations between the academia and the industry based on his research and management activities at TUAT, and has served in over 40 committees for industry and university partnership under the Japanese Ministries and their agencies.

He has received numerous awards by academic societies and governments in Japan. Included among these are: Fellow for the Contributions to handwriting recognition and its applications by IAPR in 2008; Fellow for the Development of online handwritten character recognition and its applications by IEICE Japan (Institute of Electronics, Information and Communication Engineers, Japan) in 2009; Fellow for the Research and development of user interface for pen and tablet by IPSJ (Information Processing Society of Japan) in 2014; Minister of Education and Science award for Development of handwriting-based user interface for tablet in 2016; and Contribution award for Development of tablet user interface and online handwriting recognition from Tokyo Metropolitan Government.



Publications in English

Recent Projects


Social Services

He has been serving for several committees of the Japanese government on Industry-Government-Academia partnership and those on IT-oriented and IT supported learning.