Natural language generation (NLG) is a type of artificial intelligence (AI) that generates natural language from structured data. While that sounds complicated, it simply means translating massive amounts of information into something humans can read and understand. But the data needs to be appropriately formatted for NLG to work.
An example of NLG would be translating the numbers in a spreadsheet into narratives or words to create human-readable text. NLG uses machine learning (ML), deep learning, and neural networks to make this process possible.
Read More about “Natural Language Generation”
Possibly the most successful real-world application of NLG is its use in data-to-text systems. These machines create text summaries of data in a database after analysis and processing. These textual summaries are used in decision-making.
The first data-to-text systems produced weather forecasts. The first of such deployed was FoG, used by Environment Canada to generate weather forecasts in French and English in the early 1990s. A more recent example is the U.K. Met Office’s weather forecast system.
NLG has other applications, though, as you’ll see in the next section.
Natural Language Generation Applications
NLG has come a long way since it was first used to interpret weather data. Today, NLG systems can write email subject lines, blog post snippets, report summaries, short-form ad copies, chatbot responses, and more. Here are some popular applications that use NLG:
Gmail Smart Compose
Gmail first introduced Smart Reply in 2017 so users can quickly compose responses to the emails they receive. A year later, it went a step further by releasing Smart Compose, letting users draft emails from scratch confidently. The application offers suggestions as you type, and if you like it, all you need is to click the “tab” button to use it.
Publishers like the Associated Press (AP) use Wordsmith to generate non-humanly possible pieces of content from its data for its readers. The application automatically writes earnings reports, freeing up reporters for more critical tasks requiring their attention.
Another Wordsmith user is the Orlando Magic, which uses it to generate personalized emails to spur fan engagement.
GPT-3, a language prediction model, produces coherent blog posts, press releases, and even technical manuals, often with a high degree of accuracy. Many marketers use it to turn correctly structured information into data-driven narratives. These include analytics reports, data-driven blog posts, product descriptions with standardized formats and lots of specifications, or even partial or entire articles. If you are curious about GPT-3, check out this post we wrote about it a while ago.
Alexa and Siri
Voice assistants like Apple’s Siri and Amazon’s Alexa use NLG to answer questions. Any software or device that speaks back when you talk to it uses some form of NLG.
NLG is also used heavily in many chatbots to partially or fully generate responses. More sophisticated ones use NLG to interpret what you type then create replies based on what you say.
There are many other NLG applications, of course, but the ones featured here should help you understand how it works.
How Does Natural Language Generation Differ from Natural Language Processing?
While NLG turns data into text, natural language processing (NLP) derives numbers from text.
NLG helps humans grasp numerical data faster with AI’s help. People don’t need to spend much time and effort, therefore, scouring through a massive amount of numerical data to come up with summaries that can be easily understood by all readers.
NLP, on the other hand, reads tons of human-written text (such as an in-depth research paper) and extracts critical numerical information from it. In essence, it can sum up the researchers’ significant findings faster than humans could.
You can think of NLG as a writer and NLP as a reader.
NLG and AI, in general, have certainly made life more convenient, and they will continue to do so as time passes.