Text entry and error correction on touchscreens

Today, the use of touchscreen text entry extends from short text messages to longer emails and blog posts, where the latter activities require more formal writing and text entry errors are less tolerated. Yet, many touchscreen users encounter problems during text entry. One of the challenges is the limited speed of text entry on touchscreens, which can be slower than physical keyboards. An even more notable challenge is typing errors, which considerably compromise text entry speed. My work aims to understand user interaction with touchscreen keyboards and their predictive features, especially when predictive features fail. To capture user behaviors around predictive features in standard touchscreen keyboards, I present the results of a crowd-sourced study that shows that using word prediction can reduce the number of operations per phrase but also results in a negative effect on text entry speed. To further investigate the negative effect of predictions failures on the user's mental and physical demand, performance, and effort, I present the results of a mixed-method user study that shows that the higher the frequency of failures, the more likely participants get frustrated. Also, the task is perceived to become physically more demanding with increasing failures. However, even with 20% failures, the mental and physical demand was significantly less than without predictive features. Further, to address the problem of improving text entry accuracy and speed in touchscreen devices, I present a new text entry method that offers both immediate and delayed feedback on language quality together with smart error correction features to decrease the overhead of correcting errors. The work shows that system feedback with fast correction methods can improve writing speed and accuracy. Finally, I list several design recommendations for giving users the ability to temporarily adjust the behavior of predictive features.