Automated news : 5 reasons to use it in the french regional medias

Maëlle Fouquenet
11 min readDec 13, 2019

Le Monde or L’Express, and national medias in general are not the only ones to use the automatic generation of articles on their website. The regional daily press also does it ! Le Telegramme, Ouest France, Sud Ouest, La Montagne, Nice Matin, Ebra … A lot of them have tried or prepare experiments in news automation. Not with the same means, but with rather similar objectives, and an important learning in each essay for the writing in terms of processes and methods.

Here are five good reasons for regional medias to start using automated content, other uses that are emerging in the short term, and some conditions for its success.

1- more free time for journalists
2- a comprehensive coverage
3- the creation of new sections
4 - a more regular or updated information
5- a density of contents for a better SEO

Precision first, it is possible to automate the generation of texts, including images or links, but also dynamic graphics. It is also possible to automatically generate videos.

Why using automation in regional medias

1 / Remove repetitive tasks, free up time

The first reason for automating content productions in regional medias is to free up time for journalists. By evacuating the repetitive and often uninteresting tasks of their daily work, journalists can then refocus on more fundamental missions such as investigation, analysis, reporting on the ground. Tasks that also bring more value to their media, with the goal of turning readers / users into subscribers.

Ex: the Vendée Globe race

Le Télégramme automated the race scores of the Vendée Globe in 2016.

“Historically, journalists wrote six scores a day between 6am and 11pm to give the position of each participant ,with the gaps between the competitors, in addition to the stories of the day,” says Vincent Lastennet, journalist in charge of data and innovation for Le Télégramme.

To support this work of pointing, the developer (integrated in the editorial team of the newsroom) built a robot to retrieve the positions. A text was then automatically generated with the last position of the participants, the gaps, etc. The different scenarios were pre-written by a journalist.

Initially, the texts were checked, read and titled by journalists used to write them. Then “the journalists saw that it worked and they told us that we could publish directly”. The chapo (about 400/500 characters) is then automatically updated in the CMS and published on the newspaper website six times a day, and the scores appear in the live stream.

Page dedicated to the scores of the Vendée Globe race in 2016

2 / More comprehensive coverage in few seconds

Automation makes it possible to cover a wider field, a territory in a more exhaustive way, and this, in a very fast way. Readers note that villages are no longer covered, often for lack of correspondents. The automation of certain infos could make it possible to treat again all the communes of a zone of diffusion (in particular the “infos services”).

It is a track explored by several french regional newspapers as well as Syllabs, the main provider in the generation of texts for the french media. According to Claude de Loupy, his CEO, it is possible to automate a local information feed in a municipality by integrating for example its events agenda, parties, movie schedules, the library schedules, the school year opening of registrations, the deadline to pay its housing tax, the date of the “Beaujolais Nouveau”, upcoming sports meetings … You can also choose to create a thematic feed.

Ex: the weather for all the cities

Before the automation of the weather, writing the daily web article announcing the local weather use to take 30 minutes to a journalist at Nice Matin. With automation, a different article is generated for each of the 363 communes, daily and in a few seconds. It contains the temperature and its evolution in the day, the wind speed, as well as a forecast for the next day.

“There are many villages where we never have content because we no longer have a correspondent. Result 80% of the content concerns the two or three main cities of our territory“, explains Damien Allemand, digital manager at Nice Matin. The automation of the weather makes it possible to offer an info every day, even if it is brief.

Automatically generated weather page for Menton

3 / The creation of new sections

Another possible use of automated news is to develop new topics / verticals and attract new subscribers, and new advertisers.

If we take football in Lyon, we can imagine developing a section dedicated to women’s teams (Lyon is one of the best female team of Europe) in all divisions to aim for completeness and attract fans and potential advertisers. If the journalists cover the main and local teams, they can delegate the writing of news announcements and results of matches of other teams and divisions below to an algorithm. The latter will also produce the scoreboard and the rankings automatically (we still need data). This is a content that the newspaper would not have produced without automation for lack of time and means.

This is the logic followed by the swedish regional press group MittMedia by launching two paid verticals (HockeyPuls and BandyPuls) largely powered by the automated system of United Robots (provider in which the media group has invested).

In 2017, MittMedia also launched a stream of automated articles on local real estate following a study that showed a strong interest in the local-type topic (typically, “who bought the house at the end of the street” ). The algorithm produced 25,000 articles in one year (which includes buyers’names, the price of the transaction and its location, and a photo from Google Street View). Subscribers can also receive notifications whenever property is for sale.

This topic generated 700 subscriptions. This is one of their most popular services that best turns free visitors into subscribers.

In 2018, they automated the information about business failures. In March 2019, the group reported winning 1,000 subscribers in one year across all of its 20 local news sites through automation.

Another similar example : Östgöta Media, another swedish regional press group, created in 2016 a national pure player dedicated to football under a new brand, with the aim of covering all divisions. The site is powered entirely by United Sports’s “sports algorithm”. According to Sören Karlsson (United Robots CEO), he attracted a younger audience. This has also attracted advertisers.

Klackspork is owned by Östgöta Media, which has created a new media brand.

4 / More regular coverage and/or automatic updating

The treatment of some subjects can also become more regular thanks to automation. It becomes possible to give a service information (road traffic or weather for example) every day for all cities, and even updated automatically. Which would be impossible if it was done by a journalist of the editorial staff.

Ex: fuel prices in (almost) real time

At the beginning of the “Yellow Vests crisis”, Le Télégramme created dynamic articles (which update themselves automatically) on the price of fuel for about 700 municipalities. A colossal job and impossible to perform by flesh and blood journalists because the update of texts is done every 30 minutes if a variation in price is found by the algorithm when it verifies the data on the source site.

Screenshot from the top of the page generated for the city of Brest

A text of about 1000 signs accompanies each page of commune. The robot uses the governmental database https://www.prix-carburants.gouv.fr and takes into account the petrol stations within a radius of 10km. It also generates a ranking of the cheapest stations in the area, the complete list of stations and a map.

It is a real service for their audience, ans a good reason to subscribe, since the information which is essential in a daily organization (the journeys home <> work being carried out mainly by individual cars in the major part of the territory). It responds to an economic concern also (one of the first and main Yellow Vests’claim was about the rise of gasoline price). The media then offers a decision aid (where to go fill up the cheapest possible near my home).

Vincent Lastennet also emphasizes a “good feedback from the journalists who wanted this articles to appear well on their local pages. They would also consulted them to write their articles”. Automation also helps journalists in finding information, allowing them to access it more quickly.

Ex: earthquake alerts

In the same spirit, Nice Matin is working on an automation of articles with each earthquake on its coverage area. The project is not yet complete but is inspired by the Quake Bot of the Los Angeles Times. This bot generates a draft for each earthquake recorded by the U.S. Geological Survey, as soon as it meets certain criterias (magnitude, exact location …). An editor then decides whether the information is published or not, and can possibly enrich it.

5 / More published articles and a better SEO

The automated creation of local pages (gathering statistical, tax, demographic, employment-related information, etc.) for each municipality generates a strong increase in the volume of content published on a regional web site.

This is the strategy adopted by Le Monde in 2015 to create volume and SEO before the regional election. The project was presented at a meeting Hack / Hackers in Paris in january 2016. It was for the national daily a way to “position itself on hyper local” and “promote the long tail”. Le Monde had indicated for each city page informations on population, housing, income, unemployment, community accounts, elected officials, and election results since 2002.

These newly added pages are referenced by search engines, they create a history on topics (such as local taxes or the local unemployment rate), and names of municipalities.

Highly strategic and important element for capitalizing on the 2020 municipal election.

What other possible use for algorithms in local medias ?

The thorny subject of the local correspondents

Important topic for the regional medias, the decline in the number of correspondents is a strong concern. Fewer correspondents means fewer municipalities covered. Specifically, the local correspondents provide between 50 and 80% of the editorial surface of the local and regional newspapers. But they are fewer and fewer. The temptation to find a technological solution is strong. Several regional medias focus on the subject.

During the Festival Local Info, held in Nantes in june 2019, Ouest-France project manager Michel Le Nouy ​​discussed the hypothesis of a form to be completed by local correspondents. This would have the advantage of collecting information in the form of data, while leaving the possibility of adding comments. For a football game for example, the correspondent could specify the state of the lawn.

On the other hand, difficult to automate a city council report or inauguration. “It takes repetition to make it work”, insists Claude de Loupy, from Syllabs. Without repetition to re-use paid developments, no economic interest simply.

Comments from integrated coaches directly

Swedish provider United Robots provides a program to collect feedback from coaching matches. The system generates an text message whose response is then integrated into the automated report which is itself published on the site.

Alerts with or without publication

A useful application to set up for readers and journalists is the alert function of the algorithms. During a particular event, the algorithm sends a hyper factual brief on the site, and / or on the social networks and mobile push. The alert is also sent to journalists who have no time to write the brief, essential and urgent information. They can immediately look for evidence, more accurate information, contextualize the event to produce a fairly dense first article.

The alert can only inform reporters that something fishy is going on that deserves their attention. For example, if we take the automated articles on the price of gasoline, the algorithm could be set to alert journalists (via a notification by mail or SMS) if an unusual difference (to be determined) occurs during a meeting. price variation. Following this alert, journalists decide whether or not to dig.

This system is already in place in some Swedish editorial offices in particular (via United Robots). Syllabs is also working on this subject. The alert can apply to anything that is automated from a database.

The expertise of the journalists and the team work, necessary conditions so that it works (but not only)

If many tasks seem to be able to automate the publication of information, the fields of exploration remain vast and very open, what other ideas will emerge?

For Laurence Dierickx, a developer journalist [who created a robot to collect the daily index of air quality in Brussels], no doubt: “the basis of automation is the expertise of journalists on a subject”. Indeed, journalists know where to get the info (data), they know how to gauge.

This knowledge is complemented by that of the journalists versed in the data which are also able to evaluate their quality and editorial relevance. They also know how to structure the database, and what it will be possible to do with or not.

It is the journalists who are able to explain the data, analyze and contextualize, why, the consequences of this or that information. But beware, and it’s a big challenge for the PQR, without quality data, no automation !

Finally, it is necessary to imagine which editorial product to draw concretely, which will be useful to the readers-Net surfers and which concerns them directly, and how. There, the help of a developer-in-house is valuable. And if we have graphic designers who know UX in the team, we can think of a project in a more global way.

Another important feature for developing automated content projects is the editorial creativity of journalists who open avenues for experimentation. As with the race scores at Le Télégramme, or information related to earthquakes in Nice Matin, each regional media can find automatable subjects and, ideally, that correspond to the interest of their readership and their editorial line.

If we add to this the fact that we only automate repetitive tasks (for reasons of cost and economic profitability), we are far from the fantasy that everything is automatable. Guillaume Desombres, the boss of Labsens and competitor Syllabs, even considers that “90 to 95% of the contents of a newspaper is not automatable”, because of the too specific nature of the information.

Another important element for these projects to succeed is the investment of management. It is essential that the hierarchy be involved and support the project on pain of demobilization on the ground.

Ideally, other media services should also be involved, such as marketing and sales, which could consider the monetization of a new vertical for example (which new advertisers to approach).

The media, including the PQR, have good reason to experiment, mainly to be able to make the most of its technologies whose costs have fallen sharply in recent years. The goal is not just to go faster, or to do more, it is primarily to win new Internet users and subscribers by improving the quality and breadth of their editorial coverage. An approach that relies on the editorial expertise of their teams and a managerial structure that allows these experiments.

These are essential and open up new organizational perspectives and new professions in the newsrooms. They also raise questions about media positioning in the future, including the transparency of algorithms (and therefore those used in newsrooms) and the intensive use of personal data of Internet users (which allow the personalization of home pages or reading path for example whose objectives are to generate subscriptions, but also to retain users once subscribed). And this is only the beginning of the possibilities, one of the next issues for the print media will be voice information. As for writing, the uses evolve and the technology promises a revolution to come.

--

--

Maëlle Fouquenet

Journaliste en formation/reconversion data analyst, ex responsable numérique @ESJPRO. Algo, transparence, audio, ❤#Berlin, #Nantes, #freediving et #lindyhop