Studies on the Literature
During the COVID-19 pandemic, people are using social media to obtain different types of information on an unprecedented scale. During the ongoing COVID-19 pandemic, it is necessary to understand the spread strategies of correct information, correct responses and sharing.
This study aims to classify and make sense of the concepts, labels and contents related to COVID-19 through keywords selected from among them. It is aimed to analyse Twitter data and natural language processing techniques in accordance with the concept determined here. Similar methods have been used in many studies in the literature. Based on machine learning, it is still interesting to scientists, accepting the human brain as a model and producing solutions for the problems of daily life by taking machine learning as basis. Qelik (2020) draw attention to the fact that machine learning is frequently used for problems such as prediction, classification and clustering. Algorithms such as genetic algorithms, expert systems, artificial neural networks and fuzzy logic are generally referred to as machine learning. The data related to the problem are modelled by computer algorithms for the solution and analysis of many problems.Some of the studies conducted for different fields of science are as follows: The posts made on social media accounts during the COVID-19 period in China were analysed (Li et al. 2020). Reactions to incidents, measures and strategies used to raise public awareness were modelled by machine learning method.
• In the study conducted by Koh (2020), the relationship between the quarantine applied in the COVID-19 period and “loneliness” was searched. Different findings were obtained by examining the messages with the words COV1D-19 and loneliness together.
• Fernando et al. (2019) drew attention to the rapid increase in the use of social media.
1n his analysis, findings related to the field of political science were obtained, and a proposal was made for the Large-Scale Twitter Observatory.• Sinnoth et al. (2016) focused on whether human emotions were affected by the weather. They concluded that a big data processing infrastructure, which scales to changing moods of people over time and correlates this with comprehensive, disaggregated weather data, would be necessary.
• Zhang et al. (2020) carried out a big data application in studies for businesses over a period of ten years. It has been observed that big data is still not used enough for businesses.
• Antonakaki (2020) sought to analyse online behaviour patterns, social responses, sensitivity to various entities and the nature of malicious attacks in a live network with hundreds of millions of users. Sensitivity analysis and threats such as spam, bots, fake news and hate speech were examined in the model, which they thought would help create the conceptual model of Twitter.
• Another study was carried out in the field of health by Million et al. (Million et al. 2020). A meta-analysis was conducted on the effects of chloroquine derivatives in patients based on published and unpublished reports that are publicly available on the Internet for one day (27 May 2020) for the COVID-19 outbreak. Ultimately, a meta-analysis of public clinical reports observed that chloroquine derivatives were effective in improving clinical and virological outcomes, and additionally reduced mortality in COVID-19 patients by 3 times.
• Antenuci et al. (2020) used data from Twitter to create job loss, job hunt and job posting indexes. Data were obtained by counting business-related statements such as “I lost my job” in the tweets. Social media indexes are formed of the basic components of these signals.
• In another study examining the relationship between data culture and decisionmaking performance on 108 companies in China, structural equation modelling was used to test the hypotheses.
This study contributed to the literature of big data management and governance mechanisms by establishing the relationship between decision-making performance and big data contractual and relational governance directly and through BDA capabilities (Shamim et al. 2020).• In the study conducted for Brazil by De Melo and Figueiredo (2020), approximately 4 million tweets and 18 thousand news were examined. First of all, in the study made by creating a series of keywords; users’ tweets containing hashtags, media and retweets were defined. These are also classified within themselves. All data were collected between January and May 2020. It was determined that the users make more informative postings and some drugs and active substances entered as keywords do not return much results in postings.
• Cepeda-Bolzman (2020) discussed the issue of refugees on the Twitter agenda in his study. The focus is on knowledge production and consumption. The findings highlight recurring issues related to sociocultural factors, human rights, economy news, critical thinking and religious issues. The originality of the refugee study, social sciences-related sociocultural factors derived from the tweets of minority groups were also observed.
• Using all geolocated photo tweets shared on Twitter in 2012-2013, Indaco (2020) reveals that tweet volume is a valid indicator for country-level GDP forecast. It has also shown that the geographic detail characteristics of social media posts are sufficient for predicting GDP, as specific to US cities.
• Calic-Ghasemaghaei (2020) wanted to prove the relationship between big data usage and financial performance. Whether the firms use big data to improve social performance is the research question of the study. It has been observed that these improvements are realized through organizational innovation in business practices, workplace organization and external relations.
• Durahim-Cogkun (2015) conducted user tendency and emotion analysis to determine the emotional well-being of citizens with Twitter’s voluntary information sharing structure.
In this context, they worked on a sensitivity analysis model to calculate Turkey’s Gross National Happiness (GNH). More than 35 million tweets posted in 2013 and the first quarter of 2014 were collected. The emotion analysis and the reliability of the data set were tested, and the results were found to be similar to the results of the GNH survey results of Turkish Statistical Institute by provinces and the 2013 sensitivity analysis results.• Yu-Wang (2015) study collected real-time tweets from US football fans at five 2014 FIFA World Cup matches (three matches between the US team and another opponent and two between other teams) using the twitter search API. Sensitivity analysis has been used to examine emotional responses in tweets from US football fans, particularly emotional changes (their or opponent’s) after goals. It has been observed that in the matches played by the USA team, negative emotions are fear and anger and generally increase when the opponent team scores and decrease when the USA team scores.
13.4