An algorithm capable of imitating text from a sample has been created by the University College London.
They have managed to not only teach the computer how to imitate anyone's handwriting but can also imitate the messier personalised qualities of text found in handwriting. A touch not found in the many cursive styles of handwriting with typefaces designed to replicates typical traits of real handwritten text found in word processing programs.
The three samples below are a mix of real and computer generated text. Can you tell which one was computer generated?
(Answer at bottom of page)
One practical example of how this might be useful could be when to disguise an identity on sensitive easy to recognised documents like when banks send out new credit cards.
Another example could be widespread use in the personalised gift market, a table could easily be added from a saved analysis of your handwriting and attached to a gift, bouquet of flowers or even a birthday card.
It might be an idea for the developers of this technology to not get too good at this replication. as it is at close inspection can still reveal what is real preventing any criminal forgery.
Of the above examples, it was sample B that was generated by a computer, the other two were written by a human.
01482 420150 - firstname.lastname@example.org
01482 420150 - email@example.com
Leeds Regional Office
Mortec office Park
Tel: 01133 979 567
The One Point Limited
The One Point HQ
Bridgehead Business Park
Tel: 01482 420150