The Breakthrough

Fujitsu leveraged a hierarchical model of expanded connections between neurons to better understand and interpret the features of the characters. To train the AI, Fujitsu developed tech that automatically creates a variety of character deformations so that the program wouldn’t get caught up in slight deviations.  The same team had previously achieved top-level accuracy in this field, as demonstrated by taking first place, with a recognition rate of 94.8%.

Future Implications

Fujitsu is aiming for the practical application of this technology by the end of 2015, while also further increasing the accuracy of character recognition technology and expanding its use to the recognition of media other than written characters, such as pictures and voice. The team is also looking into using the same program to interpret other languages, including Japanese.

Share This Article