When I was in high school I learned martial arts from a man who'd just arrived from mainland China. He spoke very little English and so, to better pass along the concepts and nuances of what he was trying to teach, he'd write things down in traditional Chinese.
I bought a dictionary and slowly, laboriously, translated what I could, checking back with my teacher and better refining the translations each time.
We brute forced the process.
Later, I took a couple years of Mandarin and simplified Chinese writing at University. A lot of that has faded now but I still marvel at the language all that it can convey.
That's what makes the work of Apple's handwriting recognition team — one of the many teams at Apple using machine learning to improve the quality of their products — especially fascinating to me.
From Apple's Machine Learning Journal
Handwriting recognition is more important than ever given the prevalence of mobile phones, tablets, and wearable gear like smartwatches. The large symbol inventory required to support Chinese handwriting recognition on such mobile devices poses unique challenges. This article describes how we met those challenges to achieve real-time performance on iPhone, iPad, and Apple Watch (in Scribble mode). Our recognition system, based on deep learning, accurately handles a set of up to 30,000 characters. To achieve acceptable accuracy, we paid particular attention to data collection conditions, representativeness of writing styles, and training regimen. We found that, with proper care, even larger inventories are within reach. Our experiments show that accuracy only degrades slowly as the inventory increases, as long as we use training data of sufficient quality and in sufficient quantity.
Also, and I've mentioned this before, but it's both wonderful and terrifying to me how Apple and others now refer to these processes as being trained, like pets, rather than coded, like machines.
