In the past, humans wrote product descriptions. That inevitably led to mistakes on websites, increasing the likelihood of returns. NLG has changed that.

Robert Weissgraeber, chief technology officer, AX Semantics

Natural language generation (NLG) technology powered by artificial intelligence is increasingly used across multiple industries to address some of their biggest challenges. The terminology can seem ominous, but the reality is much different: NLG is a tangible technology that can be quickly deployed in the workplace and easily grasped by employees, even those who work outside of the information technology department.

NLG is a software process by which users can transform data into a written narrative. NLG software can produce a degree of automated content, although more complex narratives require some degree of human intervention and input. The purpose of NLG is to solve some of industries’ most significant pain points. In the ecommerce sector, NLG helping to fix one of that industry’s biggest problems: return deliveries, which will cost $550 billion by 2020, an increase of 75.2% more than four years prior.

How NLG software works

The only thing companies need to get started with NLG (outside of existing computers and technology) is data. NLG software is fed unstructured data that then produces lifelike text. NLG software, in its current form, isn’t a “magic box.” It’s incapable of taking unstructured data and transforming it into legible written text without human intervention and guidance.

The applications are vast. NLG can be used to power content creation in many industries. Some of the most common implementations of NLG include business data reporting and analysis; personalized client communications via in-app messaging and email; IoT maintenance reporting and device status; individual customer financial portfolio performance updates and summaries and (more on this later), and ecommerce product descriptions and landing page content.

NLG-written narratives are designed to read as though a human wrote each one. The writing style, structure, and insights vary depending on the target audience, purpose and context of the content.

NLG can also provide data experts with an efficient way of automating the translation process that takes place when a clear, concise explanation is required to clients or co-workers who are not data experts. For example, one use of NLG is written analysis for business intelligence and analytics platforms. The process of manually writing complex reports and translating them into user-friendly text that non-experts can understand is exceptionally time-consuming and challenging. NLG enables experts to automate written analysis in language a layperson can understand.

How NLG solves an old problem 

NLG enables the user to produce thousands of unique narratives in a fraction of the time it would take someone to write them manually from scratch. Writing thousands of product descriptions is repetitive, time-consuming, and costly, whether in-house writers or freelancers are used, and add in translations, it gets expensive quickly. With NLG, retailers can promptly transform inventory specifications for every ecommerce product in a catalog into a keyword-rich description or review.


How does this assist ecommerce? It’s simple. In the past, humans wrote product descriptions. That inevitably led to mistakes. It also created language challenges if a company wanted to sell its products to shoppers outside its national borders. Another issue was that businesses were incapable of keeping product descriptions up to date and, as a result, often left incorrect or obsolete product descriptions online. When a shopper ordered a product based on that dated description and received something else, this led to a return and, quite possibly, the loss of future business from that customer.

NLG has changed all of this. First, it allows a business to quickly create and upload an entire catalog of online products while maintaining human oversight. What this means is that accurate product descriptions go up faster and yet are still unique, boosting SEO significantly. Second, since NLG works from a set of fixed data, the common errors that once ended up in descriptions are no longer a factor—unless someone introduced them at the outset. This increase in accuracy reduces returns.

Finally, when a business needs to update its entire product line, NLG provides a way to do that quickly, by providing a path for “living content” – content that updates itself automatically. Again, with correct and timely information available for each product, the return rate diminishes.

First steps 

Setting up NLG software is no longer an arduous task that requires a small army of technologists. Early NLG implementations would take months of work by professional data scientists, solutions architects, and software developers using text editing programs, such as creating a usable NLG software application. Some NLG solutions would require businesses to invest six-figure sums in an unfamiliar technology.


Modern NLG platforms are far more sophisticated, making it a simple process for a user to upload their data and begin automating their consumer-facing narratives. Most NLG software applications can now be integrated seamlessly into existing processes, too.

The result is a process that is not only easy to establish but, ultimately, provides a simple way to streamline and improve a host of business problems – and improving the return rate on ecommerce is just the start.

AX Semantics provides content-generation software.