Using AI to make journalism better. Together.



Search

Using AI to make journalism better. Together.

The Collab Challenges are a series of collaborative experiments launched by JournalismAI in 2021. They bring together media organisations from across the world to explore innovative solutions to improve journalism via the use of AI technologies. The programme was an evolution of the 2020 Collab, the first-ever collaboration of this kind.

Participants from more than 20 news organisations worldwide worked together to imagine and prototype new ideas to turbocharge journalism with AI, with the support of our regional partners: the Knight Lab at Northwestern University in the Americas, BBC News Labs and Clwstwr in EMEA, and the Times School of Media at Bennett University in APAC.

In this page, you can explore all the projects of the 2021 Collab Challenges teams and learn more about this collaborative experience directly from our regional partners:

Talking sense: Using machine learning to understand quotes

Participants: the Guardian (UK) and AFP (France)

Quotes are a key element of news articles. They help journalists explain events better and users form their own opinions. In this project, participants from the Guardian and AFP built a system that automatically extracts quotes from news articles and accurately attributes them to sources (people or organisations).

Explore the project on the Guardian's website

Watch the team's presentation at the 2021 JournalismAI Festival

Modular Journalism: An algebra for news modules to support a new kind of storytelling that is more focused on user needs

Participants: Il Sole 24 Ore (Italy), Deutsche Welle (Germany), Maharat Foundation (Lebanon), and Clwstwr (UK)

At the base of this cross-border project is the concept of modules (segments of journalistic discourse) that can be repurposed in different formats with the aid of an algorithm. The focus is on building modular-first news artefacts that are specifically created for the purpose of modularisation and show how modules can be configured and reconfigured to create stories on the same topic but meeting a range of user needs through different formats.

Explore the project at modularjournalism.com

Watch the team's presentation at the 2021 JournalismAI Festival

A journalist's guide to using AI and satellite imagery for storytelling

Participants: Bloomberg (US), Data Crítica (Mexico), La Nación (Argentina), El CLIP (Latin America)

A picture can say a thousand words. That's why this project decided to explore the use of satellite imagery and applied AI for storytelling. The primary focus has been to look at the climate crisis through this lens to see what can be reported through the observation of our planet via satellite imagery.

Find out more on the project's website

Watch the team's presentation at the 2021 JournalismAI Festival

Science in context: Generating climate fact boxes with AI

Participants: BR AI + Automation Lab (Germany) and Science Media Center Germany

News articles often lack context, which is crucial for understanding the story. News formats are concise and avoid transporting redundant information, which excludes already underserved groups of potential readers. Participants of this team decided to tackle this problem by linking concepts, terms, and entities from news articles to a knowledge graph to offer relevant context and definitions for readers who might lack some important background information.

Explore the project's website

Watch the team's presentation at the 2021 JournalismAI Festival

DockIns: Machine Learning on deadline for journalists

Participants: MuckRock (US), La Nación (Argentina), El CLIP (Latin America), Ojo Público (Peru)

The project aims to make it easy for journalists to quickly do rough classifications and sorting of large-scale document sets across a variety of languages. Building off prior document classification and reporting work at La Nación and a prototype machine learning tool developed by MuckRock, DockIns demonstrate new ways not only for newsrooms to quickly understand and map large corpus of documents, but also help automate ongoing accountability coverage.

Explore the project on MuckRock's website

Watch the team's presentation at the 2021 JournalismAI Festival

Political misogynistic discourse monitor

Participants: AzMina (Brazil), Data Crítica (Mexico), El Clip (Latin America), La Nación (Argentina)

This cross-border collaboration maps misogynistic attacks that are initiated or stimulated by political figures on Twitter. To do that, participants trained an AI model that is able to identify with a good level of assertiveness when a publication contains hate speech against women.

Explore the GitHub repository of the project

Watch the team's presentation at the 2021 JournalismAI Festival

Stimulating young people's news consumption with facts

Participants: Media City Bergen (Norway) and local partners

This collaboration between Norwegian media houses explore whether the way journalists write their articles stands in the way of younger audiences engaging with news stories and whether adding automatic "fact-finding elements" to the stories can improve on this.

Explore the project's website to find out more

Watch the team's presentation at the 2021 JournalismAI Festival

Making podcasts and on-demand radio more accessible

Participants: Sveriges Radio (Sweden)

Many journalistic podcasts have big audiences but tend to serve traditional, news-oriented audiences. By giving audiences the opportunity to find snippets related to specific topics in podcasts without having to listen through lots of other content, we can serve them better. With this project, the SR team worked on an application that segments podcasts or radio shows, transcribes the segments, tags them and clusters the tagged segments for an editor to build playlists.

Explore the project on SR's website

Watch the team's presentation at the 2021 JournalismAI Festival

Building a cross-section article network with AI

Participants: South China Morning Post (Hong Kong) and Initium Media (Hong Kong)

The project aimed to prove or disprove the theory that directing readers to articles from different sections leads to higher loyalty and subscriptions, and helps to build stronger news narratives. The team developed an AI-powered tool to recommend content based on numerous factors, creating stronger links across desks and sections.

Explore the project on the SCMP's website

Watch the team's presentation at the 2021 JournalismAI Festival

Automating storyline detection to better serve our audience

Participants: TX Group (Switzerland)

A storyline is a coherent, chronologically-ordered list of articles covering a sequence of events revolving around a particular phenomenon, with a beginning and eventually an end. Our ultimate goal is to provide context to our readers: we thus see storyline detection as an infrastructural building stone for better contextualisation. How can we use AI to automate storylines detection and better serve our audience?

Find out more about the project here

Watch the team's presentation at the 2021 JournalismAI Festival

To find out more about how the teams worked together during the 2021 Collab Challenges and what went on behind the scenes in the three cohorts, you can read these pieces by our regional partners:

Credits

Participants of the 2021 Collab Challenges presented their projects at the JournalismAI Festival. You can find all their presentations here.

The Collab Challenges are organised by the JournalismAI team at Polis – the journalism think-tank at the London School of Economics and Political Science – and supported by the Google News Initiative.

London School of Economics and Political Science

Houghton Street

London

WC2A 2AE
UK

LSE is a private company limited by guarantee, registration number 70527.

+44 (0)20 7405 7686

Campus map

Contact us

© LSE 2021