 | By Julian Park, Joyce Yanru Jiang & Anne Whitehead @ Whitehead Communications 2021 As the COVID-19 pandemic carries on into 2021, despite vaccine development, countries like Uganda continue to see their fight against the pandemic reflected in news media coverage. In 2020, Whitehead Communications applied a machine learning technique to analyse the trends in Ugandan online news coverage of the pandemic. We have now updated our study using LDA Topic Modelling to identify dominant topics of Ugandan COVID-19 coverage from March 2020 up to the end of June 2021. This analysis shows an evolution in the pertinent themes of COVID-19 coverage in Uganda as well as a changing volume of overall coverage over time. Whitehead Communications gathered 32,212 English-language news articles that include the keywords “COVID” and/or “Corona” from a selection of Uganda’s leading news websites. These websites include Uganda’s major print publications – Daily Monitor, Red Pepper and The Observer – as well as major online-only Ugandan media houses The Independent, Chimp Reports, Nile Post, PML Daily, Softpower, The Tower Post, Eagle News and Trumpet News. (Unfortunately, Uganda’s New Vision data was unavailable at the time of this analysis.) |
This research was done to broaden the chronological scope and topic diversity of previous exploratory research. As a result, the methodology of this research is largely consistent with said previous research. We used the machine learning method called Latent Dirichlet Allocation (LDA), which is a type of Natural Language Processing (NLP) that groups content into categories based on identification of keywords. As with the previous study, this research is a demonstration of qualitative methodology. It provides insights into the prevalence of the story over time and indicates key trends in how the subject of Coronavirus was covered in Uganda, though it does not provide comprehensive data on all aspects of media coverage.
PURPOSE
This research aims to expand our exploration of how COVID-19 has been covered by Ugandan online news media from the beginning of 2020 to June 2021 using machine learning. We identified and scraped articles from top Ugandan news sites publishing in English and processed them through a LDA Topic Modelling computational technique to determine which topics are covered related to the pandemic. Our research is intended to deliver further insights to parties with an interest in Uganda’s media industry, or its COVID-19 response, or the machine learning methodology as it applies to media and communications research. This update to our previous research is part of a wider research project by Whitehead Communications exploring the application of multiple new research methods to see the bigger picture, draw correlations and build stronger research-based foundations on which to develop communications strategy.
PURPOSE
This research aims to expand our exploration of how COVID-19 has been covered by Ugandan online news media from the beginning of 2020 to June 2021 using machine learning. We identified and scraped articles from top Ugandan news sites publishing in English and processed them through a LDA Topic Modelling computational technique to determine which topics are covered related to the pandemic. Our research is intended to deliver further insights to parties with an interest in Uganda’s media industry, or its COVID-19 response, or the machine learning methodology as it applies to media and communications research. This update to our previous research is part of a wider research project by Whitehead Communications exploring the application of multiple new research methods to see the bigger picture, draw correlations and build stronger research-based foundations on which to develop communications strategy.
OVERVIEW OF RESULTS
When compared to the results of our first study, the number of articles more than doubled since July 2020 (from 13,000+ to 32,000+). However, the articles per month were less than in the first few months of the pandemic, indicating a drop in interest in late 2020 and early 2021. When the first Ugandan COVID-19 case was confirmed on March 21, 2020, alongside the declaration of Uganda’s first nation-wide lockdown, there was a spike in the number of articles including mentions of COVID/Corona. April 2020 marked the peak in number of articles published through online Ugandan media in one month of 2020, which was followed by a steady decline in the following months. However, since the lowest point in January 2021, we saw a significant increase in the number of Coronavirus related articles alongside a resurgence in Ugandan COVID-19 cases and new lockdown in June 2021, setting a new all-time high for COVID-19 coverage in one month.
It is worth noting that Uganda had an election in January 2021, which was accompanied by a total shutdown of the internet for about a week, during which time online media was not able to publish as usual. This focus on a political season, along with a normalisation of the pandemic and slow progression of cases in the country, may help explain why coverage decreased between May 2020 and May 2021.
There were several recurring topics through multiple iterations of our modelling process. Topics such as “Education”, which was primarily centred around school closures and the impact of COVID-19 on students, is an example of this. Other Topics such as “Healthcare”, “Sports” and “Police Action” also appeared in almost every result in both years despite changing various parameters of the modelling process such as the number of topics. The articles that are most associated with these topics are highly distinct in vocabulary usage from the subject matter of other articles. For example, the aforementioned “Police Action” topic is associated with words such as “detain” or “arrested”. These words have a much lower chance of appearing in non-police related articles, which makes it easy for the algorithm to pick up on this pattern regardless of parameters. In addition to repeating topics, there were topics that evolved due to the additional articles used in our update.
When compared to the results of our first study, the number of articles more than doubled since July 2020 (from 13,000+ to 32,000+). However, the articles per month were less than in the first few months of the pandemic, indicating a drop in interest in late 2020 and early 2021. When the first Ugandan COVID-19 case was confirmed on March 21, 2020, alongside the declaration of Uganda’s first nation-wide lockdown, there was a spike in the number of articles including mentions of COVID/Corona. April 2020 marked the peak in number of articles published through online Ugandan media in one month of 2020, which was followed by a steady decline in the following months. However, since the lowest point in January 2021, we saw a significant increase in the number of Coronavirus related articles alongside a resurgence in Ugandan COVID-19 cases and new lockdown in June 2021, setting a new all-time high for COVID-19 coverage in one month.
It is worth noting that Uganda had an election in January 2021, which was accompanied by a total shutdown of the internet for about a week, during which time online media was not able to publish as usual. This focus on a political season, along with a normalisation of the pandemic and slow progression of cases in the country, may help explain why coverage decreased between May 2020 and May 2021.
There were several recurring topics through multiple iterations of our modelling process. Topics such as “Education”, which was primarily centred around school closures and the impact of COVID-19 on students, is an example of this. Other Topics such as “Healthcare”, “Sports” and “Police Action” also appeared in almost every result in both years despite changing various parameters of the modelling process such as the number of topics. The articles that are most associated with these topics are highly distinct in vocabulary usage from the subject matter of other articles. For example, the aforementioned “Police Action” topic is associated with words such as “detain” or “arrested”. These words have a much lower chance of appearing in non-police related articles, which makes it easy for the algorithm to pick up on this pattern regardless of parameters. In addition to repeating topics, there were topics that evolved due to the additional articles used in our update.
 | There were several topic changes between 2020 and 2021, which roughly reflected the global shift in perception about the pandemic – from an unprecedented emergency in the spring of 2020 to a continuous concern as new treatments and vaccines began to be introduced and the public and private sectors adjusted to getting on with things despite the continuing threats of the pandemic. New emerging topics in 2021 included “Travel & Tourism”, “Local Donations” and “Innovation Response To Domestic Outbreak”. The number of articles we scraped across different media houses is represented in the chart below. Daily Monitor – one of Uganda’s leading daily newspapers – produced the most online coverage among those in our sample, followed by The Independent. Other major online media PML Daily and Nile Post came in at a distant 3rd and 4th, followed by Chimp Reports, The Observer and other media houses. |
TOPICS
We tested out several parameters to determine the most effective new LDA Topic Modelling categorisation of articles. The following topics emerged in Ugandan articles about COVID-19 from the beginning (January 2020) up to the end of June 2021.
1. Healthcare
2. Vaccines
3. Economy & Finance
4. Travel & Tourism
5. Cases & Testing
6. Editorials & Personal Stories
7. President Museveni & Regional Politics
8. Electoral Politics
9. Police Action
10. Children & Women
11. Innovation Response To Domestic Outbreak 12.Sports
13. Education
14. Local Donations
15. Global Outbreak
16. Parliament, Courts & Justice
These topics indicate the main types of COVID-19 stories being published in Uganda related to coverage of the pandemic and were identified using an unsupervised machine learning algorithm, which was run multiple times with fine-tuned parameters to determine the optimal results.
Most of these topics are carried over from the previous 2020 clustering process with some modifications. However, there were some differences from last year’s assessment. One main difference was the development and distribution of COVID-19 vaccines and other treatments. Other differences indicate changes in various industries as Uganda begins to accept that life with Coronavirus will not be over as quickly as expected. This is shown in Cluster #11 where various Ugandan organisations are responding to the need for increased e-commerce and other COVID-19 caused needs. Furthermore, articles focused on travel shifted from focusing on restrictions in 2020 to integrating more stories about the re-emergence of the tourism industry in 2021 and how it has been faring with the effects of COVID-19.
More details on our methodology and an examination of key topics with representative articles are shared in the full report.
We tested out several parameters to determine the most effective new LDA Topic Modelling categorisation of articles. The following topics emerged in Ugandan articles about COVID-19 from the beginning (January 2020) up to the end of June 2021.
1. Healthcare
2. Vaccines
3. Economy & Finance
4. Travel & Tourism
5. Cases & Testing
6. Editorials & Personal Stories
7. President Museveni & Regional Politics
8. Electoral Politics
9. Police Action
10. Children & Women
11. Innovation Response To Domestic Outbreak 12.Sports
13. Education
14. Local Donations
15. Global Outbreak
16. Parliament, Courts & Justice
These topics indicate the main types of COVID-19 stories being published in Uganda related to coverage of the pandemic and were identified using an unsupervised machine learning algorithm, which was run multiple times with fine-tuned parameters to determine the optimal results.
Most of these topics are carried over from the previous 2020 clustering process with some modifications. However, there were some differences from last year’s assessment. One main difference was the development and distribution of COVID-19 vaccines and other treatments. Other differences indicate changes in various industries as Uganda begins to accept that life with Coronavirus will not be over as quickly as expected. This is shown in Cluster #11 where various Ugandan organisations are responding to the need for increased e-commerce and other COVID-19 caused needs. Furthermore, articles focused on travel shifted from focusing on restrictions in 2020 to integrating more stories about the re-emergence of the tourism industry in 2021 and how it has been faring with the effects of COVID-19.
More details on our methodology and an examination of key topics with representative articles are shared in the full report.
