4/10/2017

INJECT

 

Helping data journalists find new angles and related data.

By Claus Hesseling, data journalist and INJECT project team.

INJECT is not a data journalism tool – but it is a tool data journalists can use in many ways to improve their work. Sounds strange, but let me explain.

Data journalism often is frustrating: while crunching numbers might be fun, finding the right data sources in the first place sometimes can take a long time. Take the topic of diesel emissions for example: there are several national and international organisations collecting and processing data about this topic. While INJECT is not a search engine for databases, it can help you find a lot of fact-based news stories and uncover where the reporters got their data from. And, you might also end up with fresh ideas for sources, or for stories as well.

INJECT basically looks for related stories on the web and in the news organisation’s own archive to come up with new angles, individuals or related facts. Every time the INJECT tool retrieves news articles that might inspire a journalist to discover new angles, it also identifies the important entities. These entities can be people, places, organisations and events reported on in that story.

Building a digital creativity tool for journalists and newsrooms is the core of the European INJECT project. Instead of copying and pasting news stories from news agencies like Reuters or AP, journalists should have a straightforward way to find new ideas for stories.

Image: Screenshot of the INJECT tool.

When searching for a topic, let’s take our previous example and say “diesel” and “emissions”, INJECT offers six different dimensions for displaying results. One of the dimensions, called “backing and evidence” finds published articles that contain a lot of numbers. Data journalists might use these numbers in their reporting or get ideas for data sources they didn’t know about yet. Another dimension looks for infographics related to the search topic. Data journalists can use these to find out how other graphics departments have covered the topic.

Naturally, INJECT’s DNA is full of methods and techniques that are also used by data journalists. It uses solid natural language processing and machine learning technologies. The INJECT development team in London and Athens had to explore new frontiers during that process. For example, one of the challenges was to tweak the Named-entity recognition for various European Languages. In the first version it had to translate all indexed articles to English for processing. Now, even “smaller” languages like Norwegian can be processed without translation.

The tool is currently able to run on GoogleDocs, Wordpress, as a TinyMCE plug-in, and as a standalone web version. It searches 2.3 million news articles in almost 200 sources, in English, French, German, Norwegian and Dutch. It is also extendable to news organisations’ own archives. Indeed, the Norwegian archives are currently integrated, enabling the local newspapers in a pilot program to search through sources in their own language. Right now, newsrooms and journalists in Norway, France and the Netherlands test-drive the INJECT tool in their daily work.

The INJECT project unites 14 partners in six countries and is part of the EU commissions’s Framework for Research and Innovation, Horizon 2020: City, University of London, Bladet Sunnhordland AS, Norway; Explaain LTD, United Kingdom; Fondation Nationale Sciences Politiques, France; Hallingdoelen AS, Norway;Hordaland Bladdrift AS, Norway; The Institute of Communications and Computer Systems (ICCS), Greece; Interlink Academy for International Dialog and Journalism, Hamburg, Germany; MLabs, Norway; Rijksuniversiteit Groningen, Netherlands; Stichting VersPers Uitgeverij, Netherlands; Stichting VJ Movement Foundation, Netherlands; Universitetet i Bergen, Norway; WAN-IFRA FR, France.

Find out more and request a demo at injectproject.eu, or follow us on Twitter: @inject_en, @inject_nl, @inject_de. You can also contact us at .(JavaScript must be enabled to view this email address) or .(JavaScript must be enabled to view this email address).

Comments