Starting with Generation of Variants

Lately, we have been talking a lot about Natural Language Generation and how revolutionary it can  be for chatbot training. But, what exactly is it? In this article, I am going to explain to you the fundamentals of NLG and how quickly it can be applied to Artificial Intelligence.  The power of NLG combined with the one of AI will improve the results  achieved by the industry.

Natural Language Generation (NLG) is a computational linguistics and artificial intelligence subfield.  It is mainly  a computer system that can produce readable texts in Human Languages. 

Typically, starting from some representation of information as input, NLG systems use knowledge about language and the application domain to produce different types of texts automatically: documents, reports, explanations, help messages, and other kinds of texts.

The goal of an NLG system is to present data, or summaries of the most important aspects of the data, in an understandable form for nonexpert users.

How it was born

Much of NLP has its origins in the work on machine translation in the 1950s; to carry out machine translation, one has to not only analyze existing texts but also generate new ones. But the early machine translation experiments did not recognize the problems that give modern work in NLG its particular character. The first significant work in the field appeared during the 1970s; a problem about lexicalizing underlying conceptual material (Goldman,1974). The field took off in the 1980s with three PhD thesis of younger students McDonald (1980), Appelt (1981) and McKeown (1982).

But after that? Some little articles about NLG appeared in the pages of Computational Linguistics and other journals in the field, in some papers and conferences, but the work done in NLG tends to be underrepresented. The linguistics and NLP community have tended to present their results in a series of workshops that have sprung up in the last ten years (Kempen (1987), McDonald and Bolc (1988), Horacek and Zock (1993), Raiter and Dale (2000), Raiter (2005), Belz (2008).

What are the applications?

After a very long time, when no one seemed to be using it anymore, NLG is emerging again, and it has many real-world applications.

  • The first use of NLG was in machine translation systems, which analyze a text from a source language into a grammatical or conceptual representation, then use that to generate a corresponding text in the target language.
  •  Another early application was in expert systems, where the formal representations of rules and facts could be used to generate texts which explained the system's reasoning. Most current NLG system are used either to present information to the user, or to partially, automate the production of routine documentation. NLG can be implemented wherever there is a need to generate content from data, some of the most common uses of this system are: producing linguistic output, generating product descriptions from inventory data, creating individual financial portfolio summaries and updates at scale, business intelligence performance dashboard text explanations, real estate property descriptions and personalized customer communications.

There are many other applications of NLG but we wanted to focus more on those that can help Artificial Intelligence since it's one of the trendiest topics out there.

However, we are not the type of people who just evaluates the tools, we love to try them and that is why next week we will show you how we include NLG techniques into our current products and the result we obtain. But if you cannot wait until then try it yourself.

Download our Data Sample 

Subscribe to Email Updates

Recent Posts


see all