You’ve no doubt used the auto-spelling feature on your smartphone while texting someone. Maybe you’ve even let the predictive text feature get ahead of you, leading to a confusing message or two. Regardless, augmented writing software is what you were using, and it’s becoming more common all the time. Here’s a quick overview.
Augmented writing uses natural language processing and artificial intelligence (AI), or machine learning, to make suggestions during the writing process. It’s common today in smartphones, word processing apps such as Google Docs, and stand-alone apps like Grammarly. Augmented writing does a great job with ‘simpler’ tasks such as spellcheck and grammar, while more complex applications like job descriptions require the more robust language analytics.
What augmented writing software does
Augmented writing software relies on big data and machine learning. Over time, a computer can learn spelling, grammar, and the different parts of speech. The software can then compare something you’ve written against its data set and offer other options for you.
Augmented writing does a great job with technical issues in writing such as spelling and grammar. Comparing a word you misspell to the correct form in the data set is right up its alley. So is comparing your comma usage to what it already knows is correct.
Augmented writing software can also help to some extent with word choice and phrasing. It can offer alternative words to help make your writing either less or more formal, or to replace potentially troublesome language (e.g., curse words). It can also predict your next words or phrases based on what you’ve already typed, such as following up “I’ll see you…” with “…later.”
What it doesn’t do
Augmented writing does, however, have its limitations.
Part of that is because human language works on context, not probability, and words don’t exist in a vacuum. Take word choice, for example.
Augmented writing software uses probability to differentiate between two forms of a word. If the noun ‘building’ is more common in the data set than the verb ‘building,’ it’ll likely choose the noun. If the verb is more common, it’ll likely choose the verb. For example, it may incorrectly suggest replacing ‘I finished building’ with ‘I finished the building.’
Actually, we don’t know for certain because we can’t see into the computer’s decision-making process. With AI, the machine is doing its own analyzing and suggesting. And someone is constantly feeding it more information, so its decision-making process is constantly evolving.
Also, augmented writing mirrors whatever is in its data set. If there’s bias in the data set, for example, it’ll find its way into the software. The guidance will be a reflection of the data, not necessarily a reflection of reality.
Finally, augmented writing is somewhat binary. To use an analogy, you can ask it whether it’s sunny outside and it can tell you yes or no. However, it can’t tell you that it’s sunny but also cold and windy, so you should probably bundle up.