Click here to watch the full recording of the session

With a particular focus on Natural Language Processing (NLP), a subfield of Artificial Intelligence, InkyLab, the dedicated research and development unit of Tunisian news outlet Inkyfada, aims to create new ideas and tools to further innovation within the newsroom.

The unit has developed a range of solutions designed to help Inkyfada reach larger audiences, tailor content to platforms and user preferences, and automate time-consuming tasks for journalists, while also aiming to bridge the gap between languages in the NLP space.

“The first and most important problem that we face is the discrepancy between languages,” said Farouk Ali, Data Scientist & Head of the R&D Unit at InkyLab, during WAN-IFRA’s Digital Media Africa conference, adding that there are more than 4,500 English NLP models in existence compared to 161 in Arabic.

As Inkyfada is published in English, French, and Arabic, all solutions developed by InkyLab are made for these three languages.

Some of the projects the team has worked on include a recommendation system, a summary and title generator, as well as text-to-speech and speech-to-text solutions to automatically create an audio version of an article or a transcript of an audio file.

Helping newsrooms achieve gender balance

InkyLab has also been working with text mining, the process of extracting information from structured and unstructured content, which was used extensively in creating the Gender Tracker, a tool developed in partnership with WAN-IFRA’s Women in News programme.

Designed to help news organisations determine whether their content is gender balanced, the tool counts mentions of women, women characters, women sources, and women authors.

“It’s very accessible and easily usable, and you don’t need to put in a lot of labour to get these results and track the contents of your newspaper,” Ali said.

Concretely, news organisations can run articles through the tracker and receive a detailed report featuring information on the prominence of women and the inclusion of women’s voices and perspectives in their content. The platform also gives users the ability to make corrections in case gender has been detected incorrectly.

“The tool is always learning and always improving through the users and feedback from users,” Ali said.

Clockwise from top left, Amanda Strydom, Senior Programme Manager at Code for Africa; moderator Valérie Arnould, Deputy Director, Digital Revenue Network, WAN-IFRA; and Farouk Ali, Data Scientist & Head of the R&D Unit, InkyLab, discuss uses for AI during the Digital Media Africa conference in July.

A data platform to understand Africa’s media ecosystem

AI and other machine learning tools also play a big role in Code for Africa’s CivicSignal team.

Managed by Amanda Strydom, Senior Programme Manager at Code for Africa in South Africa, the unit comprises data analysts and technologists tasked with developing solutions for other teams within Code for Africa, as well as tracking online articles from various media sources in near real-time.

For the purpose of media monitoring, the CivicSignal team developed the CivicSignal MediaCloud in partnership with MIT, an open source and open data platform for storing, retrieving, visualising, and analysing online news.

Tracking how narratives change and evolve

Using machine learning and NLP, the MediaCloud tracks media sources in 54 countries across the continent and is able to track how narratives are evolving or being subverted.

“The free-to-use tool gives a bird’s-eye view of specific trending topics and narratives that are being pushed within certain media, and to see how they are changing, how they are developing as well as whether they are being influenced in a specific way,” Strydom said, adding that the tool also allows for comparative analysis between topics, countries and larger geographical areas.

The MediaCloud currently supports three main tools, “Explorer,” “Topic Mapper,” and “Source Manager.”

With the Explorer tool, users can track the increase or decrease in attention a specific topic has been getting, find out which key events drove coverage, and which media sources covered an issue.

Topic Mapper can dive deeper into the results of an Explorer query, offering insight into social media traction, the most influential sources that have covered a certain topic or the most relevant stories that relate to it.

The Source Manager tool allows users to browse the media sources and collections in the database.

“It’s really a tool for an investigative journalist to say ‘Hang on, what’s happening here?’ and then to be able to dig in a bit further,” Strydom said. “It really helps to provide a bird’s-eye view of the system.”