amnesty international logo 0

Amnesty International CIO John Gillespie has partnered with a London startup to help use data science to measure sentiment analysis and improve media monitoring of the global human rights charity.

The CIO of London-based Amnesty International told CIO UK that AI and machine learning are two technologies the organisation brought together to help it quickly gauge how it is being represented in the media.

The charity receives a significant amount of coverage across a broad range of topics so Gillespie and his team looked towards emerging technologies to see how to improve Amnesty's data science capabilities.

"There are many media monitoring services available, and they are great at tracking sentiment and reporting how much is being written about an organisation," Gillespie said. "This is sufficient for a company that is sending out a handful of press releases each month, but when you are issuing four or five a day and you want to know the impact of each one individually, you need something more sophisticated."

Amnesty has already started using machine learning as a research tool, including to detect and classify violence and abuse against women on social media platforms. But spending resources on investigative work is a potentially costly risk for the charity, while tracking the sentiment analysis of Amnesty on each story and the effectiveness of its press campaigns was precisely the type of large-scale, complex process that advances in data science could help with - without having any kind of negative impact on its core mission.

The organisation turned to London-based startup ASI Data Science for assistance with the process. The startup believes that AI should be accessible for everyone and organises a 'Data Science Fellowship' that enables top PhD graduates and software engineers to go through a six-week programme covering data science, data engineering and applications in industry to work on real-world big data problems.

According to Amnesty, the fellowship turned out to be a great way to experiment with data science, also proving that machine learning and AI techniques were effective tools to track the impact of each press release sent out.

Amy Bellamy, Business Intelligence Manager at Amnesty International, said: "We pitched our problem to the fellows in quite an open-ended way, we made suggestions but we were keen to let someone experiment and try different things with the data to see what was feasible."

The fellowship paired Amnesty International with Tom Begley, a University of Cambridge mathematics PhD graduate. 

Begley developed mathematical models over the course of the placement, which was used to take press coverage and determine whether each story was related to Amnesty's recent releases.

The models were then put into production, and together with a newly built interface of Amnesty's internal reporting tools were able to eliminate the tedious process of searching through news stories.

Gillespie told CIO UK how pleased he and the organisation were with the results. "The success of the fellowship project far exceeded our expectations. The opportunity for Amnesty to experiment with data science in a low cost, low-risk way was perfect for us," he said.

Amnesty's position in the industry means its concern towards how it is represented in the media is high, especially given its aim to detect and classify violence and abuse against women on social media platforms.

Gillespie was recognised in the 2016 CIO 100 for transforming Amnesty International's IT department into a reliable, modern and effective organisation, supporting the charity's regional offices and important work around the globe following a period of underinvestment in technology.