Touch keyboards have come a long way. They are not clunky pieces of software anymore, but are slick, intelligent add-ons to the idea of a keyboard. When it comes to predictive typing, we have seen phenomenal amounts of growth in quality. Most of the predictive keyboards out there can perfectly predict phrases. But SwiftKey wants to take prediction to the next level - using artificial intelligence, machine learning and neural networks.
The purpose behind what SwiftKey released today is to improve predictions and word completions to such an extent that you’ll need to type minimum characters. SwiftKey Neural is the company’s latest keyboard software that is still in its alpha stage, but promises to bring pretty amazing improvements.
What is SwiftKey Neural?
Normally, predictions are provided using the 'n-gram' methodology that works fantastic with phrases. Because people tend to use certain phrases often, the n-gram technique can learn and predict the word following the last three words. However, it is not perfect. N-gram is limited to phrases.
Human conversations happen in a completely different manner - they involve choosing words singularly from a bunch of similar words that from neural webs. These words are then knit together individually (and not as a phrase) to produce meaning. SwiftKey Neural uses this idea to predict words more correctly in the context. As I mentioned above, it uses artificial intelligence combined with neural networks to simulate the context. Here is an amazing infographic by SwiftKey that shows how it works:
You can get the SwiftKey Neural Alpha version for free on the Play Store for free.