The machine is built around glyphs, or a specific instance of a character, and can recognize a person’s specific character choices, texture, the inter-character ligatures (the joining-up between letters), and vertical and spacing. Since everyone’s handwriting has slight differences within itself, it takes the average.

Advertisement
Advertisement

Tom Haines, the lead author on the project, said that “My Text in Your Handwriting” has the potential to help people who have trouble holding a pen, or if banks or other institutions want to send out sensitive documents and want to make them look like handwritten letters.

“Stroke victims, for example, may be able to formulate letters without the concern of illegibility, or someone sending flowers as a gift could include a handwritten note without even going into the florist,” Haines said in a statement. “It could also be used in comic books where a piece of handwritten text can be translated into different languages without losing the author’s original style.”

Advertisement

As someone who signs contracts online, I can also see the potential benefits of it to better register signatures.

Of course, it can also be used to write out words in a famous writer’s penmanship, which I’m sure will make looking up quotes on the internet even more confusing and hard to check. However, study author Dr Gabriel Brostow noted that the system can actually help in detecting forgeries, since the writing can be analyzed under a microscope.

Advertisement

“We can use our software to characterise handwriting to quantify the odds that something was forged,” he said. “For example, we could calculate what ratio of people start their ‘o’s’ at the bottom versus the top and this kind of detailed analysis could reduce the forensics service’s reliance on heuristics.”

Advertisement

Researchers had participants write out certain familiar, English-language passages from the top 100 books in the Project Gutenberg database. According to the code, which was partially posted over at Github, this was done to get the most accurate sample because familiar passages are easy to write and the participants wouldn’t pause.

People were also asked to distinguish between handwritten envelopes and ones created by the software. Researchers said that people were fooled around 40 percent of the time.

Advertisement

Not sure if any software can replicate my sloppy-ass handwriting, although I’m sure that’s what everyone says. Here’s hoping my cursive can break the algorithm (it probably won’t).

Advertisement

[BBC]