We are pleased to announce that Le Temps (a Swiss daily newspaper), is one of the winners of the Digital News Initiative, a European fund launched in 2015 to support innovation in the media industry.
With EUR 45,000 in funding, our project is to develop a tool called Zombie that can identify an online newspaper’s best articles and assess when and why to republish them. Zombie will be open source for other publishers.
Le Temps publishes around 70 articles on its website every day. Some content is of only immediate interest and quickly becomes out-dated. Other content remains relevant for much longer; for instance, an in-depth report on Syrian schoolchildren or a long interview on the failings of the Swiss education system can be read again months later with the same level of interest.
When these evergreen articles are published, they are very popular. Website readers increase the level of engagement by liking, commenting on, sharing and retweeting the content. But these articles are soon forgotten because they are no longer visible on the website and can be found only by using a search engine. Every now and then, a web editor may remember an old article on a subject of newfound relevance and republish it, such as when someone famous dies. But that depends on editors’ memories – and it is simply not possible to remember all 20,000 articles that a newspaper publishes every year.
If an online newspaper could automatically find the most relevant articles that it has already published, it could republish them or repost them on social media when the time is right. These articles would then get new readers.
Zombie’s value proposition
Zombie will automatically mine all the content generated by Le Temps. It will give each article a relevance score based on average reading time, viewing history and level of engagement on social media. Using these scores, Zombie will try to match previously published articles to the day’s hottest topics. These articles could be worth republishing on Le Temps’ website or reposting on its social media pages. Zombie will send its suggestions to the Le Temps web editor and community manager through a daily email and with Slackbot.
That’s where Zombie comes in: our bot will find Le Temps’ best articles and tell its editors when and why they should republish them. Here’s how it will work:
1. Zombie will analyse articles on Le Temps’ website using data from both Chartbeat and Google Analytics. It will score each article according to its relevance and quality. This score will be calculated using the article’s reading time, viewing history and engagement on social media networks. Zombie will also identify key people, places and events mentioned in the article using semantic analysis APIs. It will create a database that, over time, will hold thousands of articles of interest that could be republished.
2. Several times a day, Zombie will see what the hottest topics are in Google Trends, Google News and Twitter’s Trending Topics. It will then check to see whether its database contains any articles related to these topics. If so, Zombie will alert Le Temps’ editorial staff in two ways: through a daily email with that day’s suggestions, and with Slack (serving as a real-time alert system).
Once alerted, the newspaper’s web editor and community manager can decide whether to republish the articles suggested by Zombie or repost them on social media.
Le Temps article entitled “Instability in cycling” was published on 27 July 2015 during the 2015 Tour de France. It was read by 50,000 people for > 3 minutes on average. It was shared 5,000 times and generated 1,800 comments on Facebook. Zombie gave the article a high score of 97.
Suppose Google Trends and Twitter’s Trending Topics bring up this year’s Tour de France as a topic. Zombie will scan its database, locate the article entitled “Instability in cycling” and then suggest it to Le Temps’ editors for republishing.
The suggestions will then be sent out by email. Here’s what the email could look like:
From: [email protected]
Subject: Today’s suggested articles for republishing
Dear web editor
Here are Zombie’s suggestions for today:
- IN THE NEWS TODAY
Tour de France (topics: cycling, Froome)
“Skinny cyclists are a good thing”, 12 December 2014 – Score: 98
“The difficult world of the Tour de France”, 11 July 2015 – Score: 97
Bombing in Iraq (topics: Islamic State, terrorism)
“Islamic State’s violent past”, 31 January 2016 – Score: 80
“Islamic State’s not finished yet”, 22 January 2013 – Score: 60
Antoine Griezmann (topics: the Euro, French football team)
“Antoine Griezmann’s home town of Mâcon”, 31 January 2016 – Score: 75
“French football’s losers are now at the top”, 31 May 2016 – Score: 45
- THIS WEEK, EVERYONE’S TALKING ABOUT:
Veganism (topics: food, lifestyle)
“So it appears I’m a flexitarian”, 31 March 2012 – Score: 73
“Should you stop eating meat?”, 27 February 2015 – Score: 45
- YOU COULD ALSO REPUBLISH:
“The generation that’s giving up possessions”, 31 January 2015 – Score: 94
“The Swiss who are helping refugees”, 13 May 2014 – Score: 91
Zombie is a solution to several problems encountered by editors: how to create qualitative reading stats that go beyond simply the number of clicks and how to improve content distribution in order to maximize impact.
We had two main sources of inspiration:
1. We got the idea for Zombie after reading an article entitled “Refreshing the evergreen” on Vox in January 2015. The article described how Vox had republished nearly 100 articles over several days as an experiment. The results showed that such evergreen articles can generate further viewing and engagement and are well perceived by readers. The Vox article concluded that in an information ecosystem inundated by content that is often mediocre, good articles remain good articles and deserve a second chance. Other newspapers, such as the Financial Times, also advocate republishing articles. But they have never put in place an automatic solution for recommending which articles to republish – which is what Zombie does.
2. We were also heavily inspired by NPR’s Carebot project. This bot provides NPR journalists with a way of qualitatively monitoring their content. The bot uses a quality indicator that is based on reading time, reader engagement on social media and an article’s completion rate. We thought this in-house indicator for measuring the quality and impact of an article was a great idea, and our relevance score is partly based on it.