Swipe Prediction and Word Frequency Analysis

[cdouble]

My experience with the swipe prediction function is sometimes fantastic but too often (and needlessly) maddening. I believe the developers have gone too far with the dictionary, capturing infrequently used words for edge cases at the expense of overall prediction accuracy.

Swipe "from" and instead I get "groin", swipe "work" and I get "whorl", swipe "weather" and I get "Starkey". I could slow down and be more precise in my swipes but but that's not the point of swipe. I would rather swipe fast for the 95% of the most common words and then add my 5% edge cases to the dictionary manually. This would vastly reduce the false positives and improve the overall experience.

So I am looking to replace the dictionary with a much shorter one based on most frequently used words. Any thoughts or advice on how to replace the dictionary file and where to find an abbreviated one are greatly appreciated.

Chris

 

[mikedt]

Might be worth reporting the problem here.
http://forum.swype.com/

Because Swype themselves would be in a much better position to fix the bug than anyone else. i.e. it's their app.

 

[cdouble]

Actually I'm using the default Samsung keyboard. They have swipe functionality similar to "Swype." I've had similar experiences with both.

Also not sure if wither of them would consider this a "bug" since they presumably made a conscious choice to use an exhaustive dictionary.

Chris