The JournalismAI team at the London School of Economics and Political Science has selected the inaugural cohort of JournalismAI Fellows, with support from the Google News Initiative: 46 journalists and technologists from 16 countries across all continents will work together in ten teams for the next six months to explore innovative solutions to improve journalism via the use of AI technologies.
Their projects will explore how a responsible use of AI can contribute to building more sustainable, inclusive, and independent journalism in all parts of the world. From building tools to support investigative journalists in their research, to automatically detecting hate speech on social media, and helping newsrooms identify underreported topics in their coverage, the JournalismAI Fellows will combine the latest AI techniques with traditional reporting methods to innovate how journalism is produced.
The ten Fellowship teams continue the JournalismAI tradition of fostering cross-border and interdisciplinary collaboration, with data scientists, reporters, product managers, researchers and software engineers working together with peers from news organisations in different continents – including collaborations between legacy brands and digital media from India and South Africa, Spain and Australia, as well as Argentina, Paraguay and the Philippines.
Due to the significant amount of exciting project proposals we received – 61 news organisations applying from 35 countries – we selected ten teams of Fellows rather than five as initially planned. Thanks to the support of the Google News Initiative and of our partners at the Northwestern University | Medill’s Knight Lab, we will guide the ten teams in the development of their projects, leading up to the presentation of their products and findings at the 2022 edition of the JournalismAI Festival.
To stay informed about the work of the fellows over the coming months, you can sign up for the JournalismAI newsletter.
The JournalismAI Fellows of 2022 are:
Our project aims to train a language model to detect hate speech on social media – in Spanish and Portuguese – directed primarily at journalists and environmental activists. The model will classify the instances of hate speech among their diverse categories and understand why and how they are happening.
Automating Visuals for Machine-Driven Content
One of the biggest challenges with AI-made coverage is the ability to generate images that enrich and compliment stories. Images are also key to successful distribution and reaching new audiences. Our project will explore a solution for automatically creating relevant imagery for content made with natural language generation.
Bad Will Hunting
Searching and comparing massive datasets for evidence of graft has become a big part of many investigative journalists’ jobs. Our project will use AI to extract NLP entities based on enriched context from long-form text, to do cross-referencing with preexisting knowledge bases/graph models. In doing so, we hope to cut down the time needed in manual curation.
We aim to create a multilingual tool for journalists that uses AI to quickly detect false claims by setting up an alert system through an interface in Teams or Slack. This system will become a central hub for newsrooms to quickly check specific claims, thereby improving accuracy and enhancing the reporting capabilities of newsrooms.
Context Cards is a machine learning model that creates and suggests context — data, bios, summary, location, timeline — to audiences and journalists, alongside an article. It will be trained on newsroom archives, and learn from editors’ feedback. We are building on prior work, including modularjournalism.com, Newscards, and Structured Stories.
🇵🇭 Jaemark Tordecilla (Editor-in-Chief and Head of Digital Media) & Raymund Sarmiento (Chief Technology Officer), GMA News Online (Philippines)
Image2Text identifies, classifies and describes video and images for newsrooms. Powered by computer vision models, it recognises objects and people in video, images and infographics, and describes them in Spanish and English through natural language. The tool seeks to promote better data governance by including perspectives from the Global South.
🇳🇴 Mads Ommundsen (Journalist & Product Owner) & Frode Norbø (Developer and Designer), Fædrelandsvennen (Norway)
Nubia is an AI-powered reporter that auto-creates development reports and data insights by transforming real time data from satellite/web camera imagery, weather and socioeconomic data into news reports, data insights and advisory that can be distributed directly to the newsroom and general audience.
Parrot is a tool and methodology to help journalists identify and measure the spread of manipulated narratives from state-controlled media. Using AI we will develop an early warning system that clusters and classifies state media generated text and then detects coordinated efforts at its dissemination.
Our project aims to help journalists to investigate influencers on a greater scale using AI techniques and developing a replicable methodology. It will track the brands, products, topics that influencers are sharing, and it will develop a scoring system to flag potential harmful or misleading content for a journalist to investigate further.
What’s there? What’s missing?
BR and MDR are both German public broadcasters with the mandate to provide multifaceted information to different kinds of audiences. We want to use NLP to build a tool that analyses our published content, as well as the reactions of our audiences, to find underreported topics in our publications.
JournalismAI is a global initiative that empowers news organisations to use artificial intelligence responsibly. It’s a project of Polis – the journalism think-tank of the London School of Economics and Political Science (LSE) – and it’s supported by the Google News Initiative.