For more than a year now, CABAR.asia’s editorial team has been effectively using machine learning (ML) techniques in journalistic reporting. What are the advantages of these methods and why should media outlets and analysts use ML in their work?
The CABAR.asia team has been using machine learning methods for more than a year. Photo: CABAR.asia
According to Abakhon Sultonazarov, IWPR’s regional director for Central Asia, he got the idea for introducing ML methods in London in 2019, during meetings with representatives of analytical centers
“At the meetings, it became clear to me that now mainly, analytics is based on artificial intelligence and machine learning. And they use these methods very actively, particularly for classified analytics ordered by multinational companies and different states,” says Sultonazarov.
According to him, in the West, even analytical memos for the leadership of countries are compiled using ML methods. Now multinational companies order classified analyses from analytical centers, for example, to decide whether to invest in this or that country. “The machine doesn’t cheat, it counts very quickly, accurately, and without emotion. It gives a clear picture and accurate analysis,” Sultonazarov outlined the merits of the ML.
According to him, the media of Central Asia do not apply ML methods in their analysis.
“It was something new for everyone. Maybe some IT companies use it, but there was nothing like that in media analytics,” Sultonazarov noted.After that, IWPR’s Central Asia office incorporated artificial intelligence and machine learning methods into CABAR.asia’s development strategy for 2019-2024 as an analytics platform.
After that, it was necessary to train editors who had no idea about ML methods. Atobek Rakhimshoev, an artificial intelligence specialist and graduate of the American University of Central Asia (AUCA), was engaged for this purpose. There were other problems, including the lack of statistical data, first of all, state statistics data, which were required for full-scale application of ML methods. This problem has been particularly acute in Tajikistan and Uzbekistan. Often, it was necessary to search for information manually, so sometimes the topics had to be simpler.During this time a series of materials were published in which machine learning methods were used in one way or another.
These materials include “Kyrgyzstan 2021: a content analysis of foreign and Russian mass media”, “Ecology through the Prism of Media in Kyrgyzstan”, “The Phenomenon of Manizha on Eurovision: Feminism Manifesto Challenging the Foundations”, “Tajik Songs: Love and Lilies Instead of Patriotic Pathos” and others. According to Atobek Rakhimshoev, an artificial intelligence specialist, he found it both interesting and difficult to work with the different editors of the CABAR.asia platform in a new and “somewhat abstract area”. Rakhimshoev acknowledged that it was not easy to find the necessary data on the Internet in Central Asia. “The second problem was that we didn’t automate the data collection process at the beginning. Later I learned how to automate the data collection process, and we solved that problem,” Rakhimshoev said.
Natalia Lee, the editor of the CABAR.asia platform, and author of the article featuring the use of ML, was “a little scared” at first to work with the new approach.
“It wasn’t clear how and in what materials to use these tools. Nevertheless, it is an interesting new experience that expands the boundaries of journalism and produces interesting results,” says Natalia Lee.
She also considers the lack of big data in Central Asia to be the major problem.
“Neither organizations nor government agencies collect large amounts of data that could then be analyzed using machine learning. Therefore, for now, the only sources for such materials in the region are the media and social media. However, there are problems with the latter because, as a rule, social networks do not allow collecting data for further processing,” says Natalia Lee.
This year she consulted an AUCA student who was an intern at IWPR and is now preparing her thesis project with a similar research methodology as the one in Natalia Lee’s article.
Marat Mamadshoev, editor-in-chief of IWPR in Tajikistan, agrees that the use of ML methods in analytical journalism has great potential.
“The pros of artificial intelligence are the ability to analyze large amounts of data very quickly and accurately. But we should not forget about the problems of AI, including errors in algorithms and models,” Mamadshoev says. According to him, few analysts in the region can successfully apply ML methods, and those who try to do so are self-taught and also in need of mentoring support. “The pace of innovation is so fast that recently acquired skills sometimes quickly become outdated. For these reasons, analysts working with artificial intelligence in their research need continuous mentoring support,” Mamadshoev believes.
Why do the media and analysts need to use the ML? And finally, a few opinions on the necessity of applying ML methods in the media.
“It seems that modern media and analytical centers (especially the latter) will not be able to produce qualitative analysis of certain socio-political phenomena in the future without understanding what insights the accumulated array of data provides. The potential of applying machine learning makes it possible to evaluate development prospects and provides predictive models, that is, variants of developments based on specific accumulated data. This is not the speculation of political scientists, not the reasoning of experts in the style of “will/won’t”, but a more accurate understanding of processes and predictive approaches to analytics. Therefore, learning to properly analyze big data is the future and partly the already realized present,” says Sergey Marinin, project manager at IWPR in Central Asia.
“The use of ML provides an opportunity to look at a particular problem from a different angle and show trends that we may have guessed about but had no confirmation,” – Natalia Lee.
“Today, the amount of available data is very large, which makes it difficult for the human brain to process all this information and creates the need to use modern analytical tools. I think it would be very useful for colleagues from the independent media to learn how to use modern tools to improve the arguments in their materials by processing large amounts of data,” – Atobek Rakhimshoev.