Nowadays, we regularly hear talk of artificial intelligence and its unlimited fields of application. An ever more innovative technology, AI continues to revolutionize our habits, both professionally and privately. Today, we invite you to discover one of its uses, NLG, and its concrete benefits for your activity.
NLG, what is it?
NLG (Natural Language Generation) or automatic generation of texts in natural language is a branch of NLP (Natural Language Processing), born after World War II.
Let’s go back a bit of time. In the 1960s, at MIT in the United States, Joseph Weizenbaum develops the first man-machine, ELIZA, a computer program capable of reformulating the statements of a human interlocutor in the form of questions. A real revolution! Today, the progress of NLP has paved the way for the development of real-world applications which we use every day. You probably already know some of them, for example the chatbots or the automatic text translation services from one language to another.
Less well known perhaps, but just as essential these days in our hyper-connected society, the automatic generation of texts makes it possible to automate the information production chain: many of the texts you read on the internet were written by robots.
How does it work? The first stage is to teach the machine to imitate human language processes. Then, based on data integrated and contextualized by humans, the program generated thousands of texts in record time.
Thus, NLG saves considerable time and productivity.
Natural language generation: why use it?
If you have followed us so far, you have understood that the possibilities of automatic text generation are multiple. And the advantages of such a solution are also numerous.
Automatic text generation: saving time and money
Writing an article can take a journalist an hour and sometimes even more, depending on the complexity of the subject and the amount of information to be analyzed. Writing a product sheet or a real estate advert can be just as time-consuming. The automatic generation of texts is therefore an ideal solution as it allows a rapid production of a large number of texts, several tens of thousands of pieces of content can be generated in only a few seconds. We often speak of ‘mass generation.’
In a few seconds, you can produce several thousands of texts which would originally probably take an hour per text if they were written by a human.
Therefore, an obvious money gain! The cost per text is considerably reduced. What’s more, this unspent time and money can be spent on high value added tasks. Thus, delegating the writing of your real estate adverts to artificial intelligence allows you to focus on your strategy to generate new clients, for example.
NLG and SEO: mass generation interest
NLG is strongly linked to SEO. To benefit from a good referencing, you must regularly supply your site with content.
That’s not all! For a good referencing, this content has to be of quality and, as much as possible, long.
The length of the texts is easily reached, since the data integrated in the algorithm are numerous.
The quality of texts, for its part, is ensured by well-established evaluation processes. This evaluation takes several forms, from scoring by an algorithm (less long and less costly but equally less precise) to careful proofreading by humans (more expensive and longer but indisputable)
Generate texts automatically…yes, but which ones?
The response to this question depends on your profile as you will have different interests depending on whether you are in the media, you are an e-merchant or a real estate agent.
Weather, match results, elections…NLG for the media
The weather is probably the first case of commercialization of automatically generated texts. From the early 1990s, FoG (for Forecast Generator), conceptualized by Environment Canada, generated marine weather reports in French and English.
NLG can equally generate sports articles (match results, prediction comments, summaries of sport issues etc.) Of course, it cannot replace the expertise of a sports journalist. Nevertheless, it allows the media to cover a larger number of competitions, including amateur matches.
Another usage: the elections. The media widely broadcasts this type of event, whether it is municipal, regional, presidential or even departmental elections. But the automatic generation of text makes it possible to publish a text by municipality! Journalists can then concentrate on in-depth articles and let the algorithm present the results of the 35,000 municipalities in France.
The list is long and it is clear that NLG can be a real asset for all media. Let’s finish with this example: the 17th March 2014, only three minutes after an earthquake near Beverly Hills, an article appears on this subject in the Los Angeles Times. Such speed was only possible thanks to the intervention of a robot journalist.
Real estate and NLG: adverts and city pages
One of the tasks the most time consuming for a real estate agent is the writing and publishing of property adverts in their charge. Whether you are selling apartments for individuals, business premises or even luxury houses and villas, your advert has to highlight the assets of your property and, sometimes, meet specific legal requirements. NLG can, from the data of these properties, generate these adverts for you. Furthermore, it can generate several versions of the same advert. All that remains is to publish them on the different websites and social media platforms.
City pages have a real importance from an SEO point of view but presenting the characteristics of a city takes time. Here too, the automatic generation of text will be a big help to you…especially since it will be easy to update these pages later. All you have to do is integrate the new data into the algorithm and generate new texts.
E-merchants: NLG and product sheets
If there is one area where data has been recorded for a long time, it is that of e-commerce. The PIM (product information management) is the proof. It is this wealth of data which makes the automatisation of product sheets easy. All of this centralized and structured data is many possible texts. Thus, we can automate the description of an air conditioning unit, as well as that of a lipstick or a screwdriver. According to the type of product, the algorithm can, in fact, identify the important data to highlight and how to analyze it.
A weight of 2.5kg would be considered as heavy for a bluetooth speaker but light for a chainsaw. On the other hand, the weight will not be a criteria of choice for the purchase of a washing machine.
Thus, even if NLG is ultimately not an extremely recent technology, it is a particularly relevant solution, at this time of referencing and the need to supply websites with content.
So, whether you are already convinced or simply curious, no matter your activity, you have understood that the automatic generation of texts is an asset for the creation of your content.
Contact us to discover how LabSense can help you to get ahead with its Scribt solution.
