The new feature is designed to anticipate what the user intends to write next and save them the effort of typing it out entirely. Microsoft’s offering joins a growing number of apps that attempt to predict your words. Predictive writing apps can be a great way to save time, experts say. “It’s a lifesaver for users when they get stuck with sentence framing,” Brad Smith, CEO of publishing software company Wordable, said in an email interview. “With text prediction, you can actually train the machine to predict text based on a user’s writing style.”
Telling You What to Write
Microsoft will roll out text prediction support to Word next month. The software reportedly will work in a similar fashion to Smart Compose in Google Docs. The features use machine learning to guess the word or phrase the author intends to type to allow faster input. “Text predictions help users write more efficiently by predicting text quickly and accurately,” Microsoft said on its website. “The feature reduces spelling and grammar errors and learns over time to give the best recommendations based on your writing style.” Users will be presented with greyed out predictions that can be accepted by pressing the “Tab” key or rejected by pressing the “Esc” key. Microsoft says that the “predictions will get better over time by learning the user’s writing style.” There’s also the option to turn off the predictions. If you’re not a Word user, there are some alternatives for text prediction software. The app Lightkey is a reliable alternative for text prediction, and it has free and paid tiers, depending on how much you or your company will use, web analyst Nate Rodriguez said in an email interview. Google’s Gmail offers text prediction, as well.
Outlook Predicts You, Too
Text prediction isn’t just for word processing. Microsoft also is rolling out text prediction for Outlook for Windows. Outlook users can accept suggestions by pressing the tab or the right arrow key. If they continue to type while ignoring the suggestions, the text predictions will disappear automatically. Outlook also can suggest responses to emails in a similar way to Gmail. When you receive a message in an email that can be answered with a short response, Outlook offers three responses you can use to reply. Microsoft also uses machine learning to help Outlook users schedule meetings. “When you view a meeting event in your calendar, Outlook might show you content relevant to the meeting, such as messages and files in your mailbox, files in your OneDrive for work or school account, or files you have permission to access in your colleagues’ OneDrive for work or school accounts or your company’s SharePoint site,” the company writes. Text prediction could be a huge help for users, experts say. “It helps with typing speed and spelling and grammar on the fly,” artificial intelligence expert Adrian Zidaritz said in an email interview. The language models allowing for text prediction have been around for some time, so why has it taken so long for Microsoft to launch the feature? “For each user, the model has to adjust to a user’s writing style so that it can gradually offer better assistance with time,” Zidaritz said. “It should work eventually like a user’s private ghostwriter.” Recent advances in natural language processing (NLP) means that text prediction could get much better in the near future, Zidaritz predicted. A sophisticated language model like GPT-3 can now write an entire story based on the title you choose. Or the model can generate one paragraph at a time or one sentence at a time. “The sky’s the limit,” when it comes to NLP, Zidaritz said.
