key: cord-0201024-qxtu5z49 authors: Saire, Josimar Chire; Panford-Quainoo, Kobby title: Twitter Interaction to Analyze Covid-19 Impact in Ghana, Africa from March to July date: 2020-08-27 journal: nan DOI: nan sha: a265e6978dad9cd52b0344546b76ca39eab6ef41 doc_id: 201024 cord_uid: qxtu5z49 The novel coronavirus, COVID-19, has impacted various aspects of the world from tourism, business, education, and many more. Like for every country, the global pandemic has imposed similar effects on Ghana. During this period, citizens of this country have used social networks as a platform to find and disseminate information about the infectious disease and also share their own opinions and sentiments. In this study, we use text mining to draw insights from data collected from the social network, Twitter. Our exploration of the data led us to understand the most frequent topics raised in the Greater Accra region of Ghana from March to July 2020. We observe that the engagement of users of this social network was initially high in March but declined from April to July. The reason was probably that the people were becoming more adapted to the situation after an initial shock when the disease was announced in the country. We also found certain words in these tweets of users that enabled us to understand the sentiments and mental state of individuals at the time. The global pandemic, COVID-19, formally known as the severe acute respiratory syndrome coronavirus 2 (SARS-COV-2) [1] , was declared a pandemic and a global threat by the World Health Organization (WHO) on 11-03-2020 [2] . Since December 2019, when its outbreak was first announced in the city of Wuhan in China, the disease has spread very rapidly to many other countries. As of 11-03-2020, when the disease was declared a global pandemic, a total of 6,546 cases and 355 deaths had been recorded by WHO [3] . Ghana had its first imported case of COVID-19 in Accra, the nations capital, and the regional capital of the Greater Accra Region, on 12-03-2020 [4] , [5] . The disease then rapidly spread across the other regions. As of 3-08-2020, 10:35 GMT+2, a total of 37,014 cases had been confirmed. The Greater Accra Region alone accounted for about 18,882, which is about 51% of the total cases in Ghana [6] . Figure 1 summarized the events from Ghana, etween March-July. The coronavirus disease is caused by a pathogen that attacks the respiratory system of humans. Non-infected persons may contract the disease after physical contact with an infected person or contaminated surfaces and objects. The disease is characterized by fever, dry cough, and tiredness, which manifests within 14 days of infection. Other symptoms include diarrhea, sore throat, conjunctivitis, headache, and loss of taste and smell [7], [8] . To slow down and control the spread of the virus, The WHO defined some measures in the absence of an approved cure [9] . Many governments soon adopted and implemented these measures in their own countries. These measures include physical distancing between persons, wearing face masks, avoiding physical interactions like handshaking, and hugging, and lockdowns, etc [10] . These measures have been communicated to individuals through various communication channels, including social media platforms like Twitter, Facebook, etc. As a result, concerned individuals and public health officials are able to stay up to date with research findings, statistical and precautionary measures. Others use these platforms to share their own opinions and sentiments towards media reports or news [11] . A careful analysis of what individuals talk about will help satisfy digital epidemiological needs for efficient disease surveillance and case severity analysis. Healthcare providers or policymakers will then understand the state of the situation, thereby making it easy to reach out with available remedies [12] . In this work, we use data sourced from Twitter, which is the most used social network after Facebook and YouTube [13] by a section of the Ghanaian population in the Greater Accra Region. We contribute by providing a data-based assessment of the COVID-19 situation and the populations response. This will aid healthcare providers and educators to assess how individuals at the target location feel, respond and understand matters concerning the COVID-19 pandemic. Based on this information, healthcare needs and decisions that are more suited to the population can be made and implemented. This section describes the proposal to analyze the situation in the country of Ghana in detail. The next step are part of the proposal: -Select keywords related to covid19 pandemic -Set the parameters to collect related data -Pre-processing text -Visualization to support Analysis The scope of the analysis is the Greater Accra region of Ghana. This choice was informed by the concentration of the population and access to the internet. The search terms used to retrieve information from the social network are: -'covid', 'coronavirus', 'sars' Twitter is a social network with thousands of users sending micro-messages, up to 240 characters. Then, the collection process is through Twitter Search function using the parameters that follow: date: 01-03-2020 to 15-07-2020 terms: the chosen words mentioned in previous subsection geolocalization: 5.646129,-0.065209, see Fig. 2 language: English radius: 25.3347 km In this step, the following procedures are taken to remove unwanted words and characters while maintaining the relevant ones. We also create visualisations to support our analysis and visual understanding. The last step is useful to avoid coronavirus keywords and discover other terms related to the topic. According to the objective, this set of words can be chosen. After cleaning the data, the text is ready to create graphics to help the understanding of the situation in Ghana during the last months. -Bar plots to show the frequency of tweets per day, per month -Cloud of words to visualize the most frequent terms per month -Line plots to show the trend of data This section explains the results obtained and the respective analysis. First subsection 3.1 introduces the description of the dataset. Next subsections 3.2, 3.3 presents the frequency of post daily,monthly and 3.3, the most frequent terms per month. Finally, a crossing graphic is presented in 3.5 to analyze the relationship between Twitter, Google Trends, and new reported cases per day. This dataset is the result of the collection explained in 2.2 previously. -Total size of dataset: 9475 tweets -Fields: date(YYYY-MM-DD), text(alphanumeric) -Range Date: 01-03-2020 to 15-07-2020 -Language: English First, it is important to know how many tweets/posts related to the topic of the study because this can express the interest of the users about the topic around covid-19 pandemic. Figure 3 , shows March and April had the highest number of posts and from April there is less tweets month after month. April 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 0 50 100 150 200 March 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 0 25 50 75 100 125 150 175 April 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 0 20 40 60 80 100 May 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 0 20 40 60 80 June 01 02 03 04 05 06 07 08 09 10 11 The highlighted dates can be the start of an elaborated study around these days, then the study can be extended. The terms: covid19, coronavirus, sars were extracted because the intention is to know the most frequent terms around these keywords per month. First, graphic 5 shows case, people, pandemic, update, positive. Then, this keywords are more related to Public Health reports. Next figure 7 presents the most frequent N-grams with N=2,3 to perform a better analysis about the previous terms. By consequence, it possible to confirm most of the tweets are related to actions of government, i.e. president akufo addo, stay home, taken combat spread, update measure taken. The choice of words of users on social networks indicates their attitude and feeling about the situation they post about. In this section, we focus on words that conveys sentiments of the user from their composition. We categorized the pre-processed tokens into three namely; positive, negative and neutral based on their polarity scores. In figure 8 , frequently occurring words across months are "death", "die", "kill". These words suggests that people were worried about the alarming death rate of the pandemic and the cause of the panic among them. Again, "racism" appears in May and more frequently in June pointing out another cause of distress that likely surrounds the issues and behaviours against racial minorities. April May June July First, Twitter can be a useful source of data. Process to analyze text includes the cleaning of the text. Second, there was an increasing interest on the pandemic as topic in Ghana user at the beginning, during the first reported cases. But, during the time, the interest about the pandemic were decreasing. The evidence for this affirmation is found in Google searches, Twitter posts daily, in spite of the number of new cases still is growing up. Covid-19 and the cardiovascular system World health organisation (who) coronavirus disease (covid19) outbreak WHO: Who health emergency dashboard Fighting covid-19: Fear and internal conflict among older adults in ghana Global stability and cost-effectiveness analysis of covid-19 considering the impact of the environment: using data from ghana GHS: Regional distribution of active cases 7. WHO: Q&a on coronavirus Sars-cov-2 induced diarrhoea as onset symptom in patient with covid-19 Covid-19 strategy update Limiting spread of covid-19 in ghana: Compliance audit of selected transportation stations in the greater accra region of ghana Communication related health crisis on social media: a case of covid-19 outbreak Characterizing the propagation of situational information in social media during covid-19 epidemic: A case study on weibo GlobalStats: Social media stats in ghana -july 2020