Skip to main content

Predictive Text Simplifies Your Life

Text prediction is unknowingly an important part of our lives. With predictive text, we can utilize shorthand scripting (writing only part of the word that we are intending to write) much more than in the past, since the predictive text can finish the words for us. The basic flow of how predictive text works is the program first “compares each character in a candidate word to a character in a corresponding position in the input sequence.” Then, “Calculate a score based on the character comparison. If candidate word and input sequence is different in only one character position, add a first bonus to the score.” Finally, “If a candidate word and input sequence is different in only one set of transposed characters, add a second bonus to the score,” (Henry). The program running the prediction can become more accurate the more it is used and corrected by the user. At first, the predictions may be funny and very outlandish than originally intended, however, as a user provides more feedback to the program (keyboard), the program can rank words more effectively. For example, the more a user ignores a suggested word, the less frequently the program would think that you are intending to write that word in the future.

Why would we want to use up more storage and computing power for our keyboards? Because it makes our lives easier. The more we use our predictive text, the more the program’s word ranking system becomes effective to us. Each person’s predictive text program is tailored towards their writing styles. Some people may use the word bot more often than boy. In some situations, the distinction between two words may be difficult to distinguish, i.e. “I don’t know if it was the boy or the dog” vs. “I don’t know if it was the bot or the dog”. However, as a user continues to choose bot over boy, the program will begin to add more “weight” (or a higher score) to the word “bot”, so that the next time a user is writing, when they type b, the text prediction would first suggest bot. Comprehensive programs will learn to read what is being typed and compare it to the data that it has collected from previous uses of the keyboard. As Henry states, some of the best programs can predict full sentences just from what a user selects as the first word. This is from their ability to learn from what a user types into the keyboard, and the program continually refining the weighting system.

This article touches on how users can shape the way that these predictive text programs rank the words that they suggest. By “training” the programs, users can create a keyboard that is more likely to predict what they are trying to say, before they even need to type the words out. Like how Google decides which advertisements to show a user when they search a phrase, or which links to display first, predictive text also ranks the best fit for each situation. Whoever developed the keyboard that you use is betting that you’ll continue to use their program because their keyboard simplifies your life the most. The less you write, the more it works.


Henry, Alan. “How Predictive Keyboards Work (and How You Can Train Yours Better).” Lifehacker, 8 Oct. 2014, 11:00 AM,


Leave a Reply

Blogging Calendar

October 2017
« Sep   Nov »