tech

Fujitsu develops accurate handwritten Chinese character recognition technology

7 Comments

Fujitsu Research and Development Center Co Ltd and Fujitsu Laboratories Ltd have developed the industry's most accurate handwritten character recognition technology that employs artificial intelligence (AI) to mimic the workings of the human brain.

Conventional handwritten character recognition attempts to identify a character based on the number and angle of its strokes. Severely misshapen characters, however, might not be recognized using this method, and increasing recognition accuracy requires the character to be "learned" in a time-consuming process. This new technology uses AI to recognize character features, in the same way a human brain does, to learn more quickly and in greater detail. This slashes character learning time to 1/17th previous levels while achieving recognition accuracy of 94.8%.

This technology came in at first place in a recent contest of handwritten Chinese character recognition, the results of which are to be officially presented at the world's largest international conference on document image processing, the International Conference on Document Analysis and Recognition (ICDAR 2013), opening Aug 25 in Washington, DC.

Fujitsu said the technology promises to greatly improve the efficiency of data entry when digitizing handwritten text.

© Japan Today

©2021 GPlusMedia Inc.

7 Comments
Login to comment

This is amazing!

0 ( +1 / -1 )

If previous methods also "learned" wouldn't they be using AI as well?

1 ( +1 / -0 )

The headline says that Fujitsu have developed an "accurate" character recognition technology implying that others are inaccurate, however its if you look a the competition report on the ICDAR 2013 website the competition is split into 5 tasks and Fujitsu performs best in only one of the tasks. For this task (offline character recognition) Fujitsu are 94.77% accurate and next system is 94.42% accurate. So the headline implies that accurate is somewhere in between these two figures.

The article makes it sound like operating "the way the human brain does" is somehow novel but in fact they're using neural networks which have applied to character recognition for donkeys years.

I'm not saying this lab in Beijing hasn't done some impressive stuff here, just that the article in my opinion seems misleading.

2 ( +2 / -0 )

The free Android Kanji Recognizer software is quite accurate, and was of help in many cases... It can't identify characters written cursively though (as my handwriting is quite cursive, with the number of strokes greatly decreasing)...

0 ( +1 / -1 )

By Chinese characters do they mean the Simplified Chinese that is used in China or Japanese Kanji which is mostly traditional? Or both? One of my electronic dictionaries does a fair job of recognizing kanji; even if fails to recognize the character it suggests other ones that are similar in form and/or stroke order.

1 ( +1 / -0 )

By Chinese characters do they mean the Simplified Chinese that is used in China or Japanese Kanji which is mostly traditional?

To a computer they would make no difference. It's just a matter of tweaking the software. I just wonder if it can really comprehend my juvenile scrawl.

0 ( +0 / -0 )

According to the competition results published at ICDAR 2013 site, the system developed by Fujitsu's Beijing R&D center yielded the most accurate results in 1 of 5 tasks only. This task was "offline (image-based) recognition of isolated characters" and the system developed by The Dalle Molle Institute for Artificial Intelligence (IDSIA), Switzerland achieved almost same accuracy. The School of Software at Harbin Institute of Technology (HIT) was the leader in 2 of 5 tasks.

http://www.nlpr.ia.ac.cn/events/CHRcompetition2013/competition/ICDAR%202013%20CHR%20competition.pdf

Quoting from this paper: "The performance of Fujitsu and IDSIAnn ... are by far inferior to human recognition performance."

0 ( +0 / -0 )

Login to leave a comment

Facebook users

Use your Facebook account to login or register with JapanToday. By doing so, you will also receive an email inviting you to receive our news alerts.

Facebook Connect

Login with your JapanToday account

User registration

Articles, Offers & Useful Resources

A mix of what's trending on our other sites