Are there any up-to-date works reporting how efficient word prediction and spelling correction systems are in terms of user experience?

Question

I'm looking for both methods for measuring efficiency of such systems, and any empirical studies which I can use as an evidence to convince investors. For example:

• 25% of Android users use word prediction module every day.
• Only 10% of spell correction suggestions of Android is accepted by users.

Answer 1

I haven't read this paper through properly but it (or some of the references it cites) looks like it will contain the information you are looking for: The length of text messages and use of predictive text: Who uses it and how much do they have to say?