Search
  • Petch

How Autocorrect Works

We’ve all complained about autocorrect before to our friends, tired of having to correct autocorrect for its corrections. There’s even a myriad of memes on Google making fun of autocorrect. Yet, this causes the benefits of autocorrect to be overlooked. In most cases, autocorrect amends the text such that each misspell and mistype becomes nonexistent.


How It Works

Autocorrection software is quite simplistic in nature. To correct things like spelling mistakes, it checks each word typed against a list of words stored in a comprehensive built-in dictionary. If there are no matches, it then suggests alternatives using contextual factors to ensure accurate words are proposed. For example, since phone keyboards are small, autocorrect often suggests words containing letters close to the letters the user had previously typed.



Otherwise, such software also predicts the next word you are going – this is known as autocomplete. This uses more complicated AI and machine learning technology that systems like iPhone and Android are familiar with. Essentially, contextual mapping is used whereby the AI calculates the probability of what the user meant to type. The core of this is natural language processing (NLP) and neural networks.


Improving Autocorrect

Autocorrection systems are becoming increasingly advanced and sophisticated. Notably, larger dictionaries are being used alongside the addition of colloquialisms to optimise the user experience for mobile users. With the evolution of language, it is important for the dictionaries to be regularly updated. Dictionaries have also been expanded to incorporate your data.



Sources

​​https://slate.com/technology/2010/07/how-your-cell-phone-s-autocorrect-software-works-and-why-it-s-getting-better.html

https://www.typewise.app/blog/build-autocorrect-emoji-text-prediction-ai


Written by Nichapatr (Petch) Lomtakul