1. Introduction

In 2020 the world is hit by the global pandemic known as COVID-19. In this sense, for Guo et al. (2020) COVID-19 has been characterized by respiratory difficulties, and community transmission has been accelerating around the world (ECDC, 2020). The COVID-19 pandemic is changing human interaction (Kye and Hwang, 2020), generating business crises (Crick and Crick 2020), economic crises (Nicola et al., 2020), crises and risks in the family (Campbell, 2020), changes and crises at work (Bartsch et al., 2020).

It should be added that crisis communication is essential to reduce uncertainty, maintain the company's reputation and avoid negative consequences. In this framework, social media has emerged as a crisis communication research area (Zeng and Gerritsen, 2014) and Twitter has been the main means of crisis communication during the COVID-19 pandemic.

This research aims to examine response strategies to crisis communication from companies and organizations during the COVID-19 pandemic regarding perceived risk and crisis, through the main topics of discussion in Twitter hashtags about the COVID-19 pandemic during the COVID-19 pandemic regarding perceived risk and crisis, through examining the main topics of discussion in Twitter hashtags about the COVID-19 pandemic and contribute to the literature gap that exists about how companies and organizations carry out their response strategies to crisis communication to avoid negative consequences.

2. Literature review

Crisis communication

A crisis is an organizational or business problem that may have natural causes or man-made disasters that threaten the operation of a company, its reputation, and its ability to do business (Weiner, 2006), crises are sudden and unexpected and pose both a financial and reputational issue that can disrupt company operations and involve a wide variety of stakeholders such as the community, employees, customers, suppliers, and shareholders (Coombs, 2007).

Specifically, COVID-19 has generated a global crisis (Maak et al., 2021) and global crises are characterized by the struggle for the survival of countries, companies, and citizens in the face of these crises, companies have adopted commercial initiatives aimed at preventing the loss of customers, the decrease in income and the erosion of market share. (Salunkhe et al., 2021). During the COVID-19 crisis, people paid more attention to messages related to COVID from companies on the social networks Facebook and Twitter than to any other type of messages and acted by commenting and sharing with other users of said networks (Kwok et al., 2021), this crisis communication on Twitter also includes government and health agencies (Watkins and Clevenger, 2021; Wang et al., 2021).

It is important to mention that crisis communication often transfers to a great extent the risks perceived by people, and in times of pandemic it was no exception, so it is important to highlight how the literature shows various dimensions of perceived risk during the COVID-19 pandemic, there was a perception of financial risk in clients in the hotel industry (e.g. Quan, 2022), perception of health risk in tourism industry (e.g. Bae and Chang, 2021; Rather 2021), and in hospitality industry (e.g. Leung and Cai, 2021), perception of economic risks for customers and businesses due to confinement restrictions (e.g. Foroudi et al., 2021) perception of risk in families by contagion in health workers (e.g. Gorini et al., 2020), perception of social risk due to social distancing and isolation (e.g. Eddy, 2021), perception of risk in businesses due to decreased sales because the restrictions and change in consumption habits in COVID-19 pandemic, for example in restaurants, hotels and leisure industries (e.g. Parady et al., 2020), perception of risk in government intervention in actions against COVID-19 (e.g. Duan et al., 2020), perception of risk at work due to the probability of contagion and job instability (e.g Han et al., 2021).

It is worth emphasizing that crisis management has as its main axis the adequate crisis communication strategy between the interested parties and these strategies depends on the type of crisis, the crisis, and the organization's responsibility for the crisis (Coombs, 2015). The theory of situational crisis communication tries to establish crisis response strategies with positive results for the organization in the public perception of the crisis and the attitude towards the organization to protect its reputation and reduce the negative effects (Coombs, 2007; Coombs, 2014). Likewise, crisis communication is crucial to reducing perceived uncertainties during the crisis (Charoensukmongkol and Phungsoonthorn 2022), and managing reputation (Christensen and Lægreid, 2020).

On the other hand, in situational communication crisis theory there are three primary crisis response strategies: (a) deny crisis response strategies, this strategy completely reallocates blame away from the organization, crisis denial strategies should be used in victim crises when the organization is faced with rumors or accusations that are damaging, but not true, (b) diminish crisis response strategies, this strategy works to minimize the amount of responsibility placed on the organization minimizing the perceived damage caused by the crisis, diminish crisis strategies should be used more frequently in response to victim crises where the company is not to blame for the problem, and (c) rebuild crisis response strategies, this strategy aims to rebuild relationships with stakeholders, this is achieved by taking responsibility for the crisis and offering apologies or compensation to those affected by the outcome (Coombs, 2007).

The networked crisis communication theory considers that crisis communications distributed by social media can provoke different responses, which are affected or impacted by the medium used, the type of crisis, and people's emotions (Schultz et al., 2011; Utz et al., 2013). Is little known in the literature on the effect of crisis communication in social networks from the perspective of the recipient since social networks can facilitate responses from multiple voices of organizations and the public or consumers (Liu et al., 2016; Utz et al., 2013).

Understanding online dialogues would help to understand communication in crisis (Lin et al., 2016). Particularly, the literature points to how economic actors develop and carry out their constructions of reality (make sense) to maintain control over the situation and institutional changes (Schultz & Raupp, 2010) and to maintain control over the situation and institutional changes (Wu et al., 2016).

Specifically, the literature is clear that in a communication crisis, companies behave differently in a group setting compared to when they face a crisis individually (Comyns and Franklin-Johnson, 2018) being that many times there is a taxonomy of crisis messages in social networks that interested parties can send to organizations (Roshan et al., 2016). As well as, in which public entities communicate with each other rather than with organizations about disasters and predict a wider variety of results of crisis communication (Liu et al., 2016). It seems important that government organizations emphasize providing instructive information to their primary audiences, such as guidelines on how to respond to the crisis using social networks more frequently than traditional means of responding to crisis (Kim and Liu, 2012) using the strategy of briefing information predominantly before and during a disaster, while information adjustment and reinforcement strategies are used more during post-disaster recovery (Liu et al., 2018).

Crisis communication and social media

Twitter is a social network that has allowed dialogues and reactions in communications in crises, becoming a functional information center during disasters (Schwarz, et al., 2016). Twitter is one of the first sources of information to obtain information (Barbe and Pennington-Gray, 2018). Therefore, Twitter has been more effective than Facebook and Instagram in terms of restoring a company's reputation after a crisis, as it improves behavioral intentions and triggers positive engagement (Triantafillidou and Yannas, 2020). For this, the tweets will recycle links to news sources inherited limitations that plagued media coverage in their inability for quick response, as well as defensive and offensive tactical dialogues, such as increasing organized events, outreach, and expert opinion statements (Roundtree, 2018).

I would have to say then that, in social media, the approval or disapproval of other people or the way other people are perceived are social norms that influence people's behavior (Cialdini and Goldstein, 2004). Today, social networks make people's behaviors visible and extend them through harmful behaviors or behaviors that benefit, through communications that can be extended to networks of a friend, friends of friends, and even friends of friends of friends (Christakis and Fowler, 2013).

For Roundtree (2018) , tweets will recycle links to news sources inherited limitations that plagued media coverage in their inability for quick response, as well as defensive and offensive tactical dialogues, such as increasing organized events, outreach, and expert opinion statements. Adicioanllity, Lee (2021) has highlighted how emotions of various kinds have recently had a direct impact on decisions, and resource management.

It is important to mention that in the institutional part, social presence strategies were effective in generating public attention on Twitter. Also, express appreciation, provide guidance, and inform the public about the actions of governments with public attention (Mazid, 2022).

Thus, for example, Aldekhyyel et al. (2022) mention how communications from government entities on issues related to COVID-19 published by Saudi government ministries on Twitter can be classified into those related to the disease and those not related to the disease. For the authors, the integration of behavioral theories in the development of health risk communication should be taken seriously by government communication specialists who manage social network accounts, since these theories help to underline the determinants of people's behaviors and their reactions.

Carvache-Franco et al. (2022) in their study analyzed the discussion topics in the popular Twitter's tourism hashtags during the COVID-19 crisis related to coastal and marine tourism, their findings contribute to the relationship of the topics with the motivational dimensions of the tourist, and the discussed destinations on marine and coastal tourism in times of crisis due to the COVID-2019 Pandemic in Twitter tourism hashtags, and companies will also be able to improve their communication strategies and develop post-pandemic products, and in the study of Carvache-Franco et al. (2022) stand out as Twitter became a medium of communication during pandemic crises, helping to reduce negative perceptions and harmful effects of tourism crises in companies and destinations.

On the other hand, the motivations and behaviors found within each of the public in crisis -- influencers and followers- are fundamentally different from each other by nature throughout the regenerative crisis situated (AO & Mak, 2021). Likewise, the importance of the influence of third parties in crisis communication and the need to use both social and traditional networks in response to crises (Austin et al., 2012) the medium matters more than the message (Schultz et al., 2011).

We must add that confidence in crisis communication is affected by the characteristics of communication, transparency, empathy, opportunity, quality, and this communication in social media occurs with various actors including public health, media, and the public who are co-producing and responding to the message, so crisis communication should be monitored to assess the acceptance of the message (MacKay et al., 2021). During the pandemic, changes in governors' communication patterns were accompanied by changes in public perception of governors' responses (London and Matthews, 2021).

Among the characteristics of this crisis communication carried out by companies on Twitter during the COVID-19 pandemic, the main concern is that of products and services, and in which their main strategy to address this concern was the exchange of information (Chong and Momin, 2021) and the response to the crisis communication of companies is that consumer confidence has been recovered when this communication has been emotionally directed (Wang et al., 2021). The strategies used by European airlines during the COVID-19 pandemic are mainly to instruct and adjust crisis communication, deny, diminish, rebuild, bolster and Twitter replaced Facebook as the main communication channel in the crisis (Scheiwiller and Zizka, 2021).

In crisis communication, relational communication seems to be more effective than informational communication (Einwiller et al., 2021). It is very clear that social networks for companies are effective means to communicate with their customers with negative emotions in a pandemic crisis such as COVID-19, what remains unclear is how companies should respond to social networks in the face of a crisis pandemic to avoid negative consequences and it seems that the strategies that have worked best in Twitter are those that evoke positive emotions among consumers (Li et al., 2021).

In crisis communication during pandemic COVID-19, evidence has been found about diminished crisis response strategies in EU airlines with evidence that the strategy used has a different application between classic airlines and low-cost airlines (Chmielewska-Muciek, et al., 2021), other evidence that exists in airlines is the use of five different crisis response strategies as instructing and adjusting information, deny, diminish, rebuild, and bolster (Scheiwiller and Zizka, 2021) and other airline evidence reports the use of rebuild crisis response strategies (Ou and Wong, 2021) while in universities has been reported with positive results in building reputation (Dominic et al., 2021) and evidence how corporate crisis response strategy and response using rebuild consumer trust (Wang et al., 2021).

Likewise, government agencies have promoted, for example, communication on issues associated with confinement, online learning, digital platforms, empowerment, responsibility, reports not related to the disease, local and international news, and general statements (Aldekhyyel et al. 2022).

Additionally, in Tourism offices in Spain and Italy, for example, the types of content related to the category of communication on the Covid-19 crisis on Twitter, focused on communication of tourist efforts to face the Covid-19 crisis or communications to safeguard the image of the destination as a tourist attraction (Huertas et al., 2020).

Twitter data processing allows the identification of the main topics mentioned by tourists and in many cases, the emotional expressions they have can be easily detected (Park et al., 2016), which the above leads directly to a better understanding of the attitudes and behaviors of tourist experiences, especially in the motivations of the tourist (Gilbert, 2016; Park et al., 2016). Specifically, we will mention that tourist behavior and the purpose of the trip can be inferred through Twitter data (Nozawa et al., 2016) since travelers have different reactions to social media information about travel planning (Zeng and Gerritsen, 2014).

As a result of the above, Twitter allows obtaining the consumer's perception, by identifying popular words and emotions expressed by tourists in the comments of this social network (Park et al., 2016). Consequently, by analyzing Twitter data, customer preferences can be identified, which directly generates support for companies of different services when planning the services, they offer to their customers (Sinha et al., 2018).

Crisis response strategies represent the words or terms and actions that organizations can take during a crisis (Coombs, 2007), and in the literature exists a gap about how companies and organizations carry out their strategies in response to crisis communication to avoid negative consequences. For the above, text analysis can be used as an exploratory analysis, but it can also be used to test hypotheses or interrelationships between constructs (Harlow & Oswald, 2016). This study asks the following research question: what response strategies from companies and organizations represent the topics discussed about risk and crisis in communication in Twitter COVID-19's hashtags?

3. Methodology

Twitter data provides a means to analyze the attitudes and behaviors of a broad spectrum of the population (Harlow and Oswald, 2016). Through the process of Twitter's big data, it is obtained data patterns and text sentiment analysis which is an important twitter data processing technique (Kirilenko et al., 2018).

Word association

The term association is a technique for finding syntagmatic relationships (syntagmatic relation) between terms or words when they appear together in a context (Correia et al., 2018). Through the analysis of the frequencies of words and hashtags, relevant topics and keywords for tourism are identified (Park et al., 2016). In the association of terms, knowledge is derived from patterns and relationships that can be used to reveal facts, trends, or constructs.

The association of terms generally uses a quantitative approach to approach the analysis of larger volumes of texts and helps to discover knowledge by increasing the volume of the text to be analyzed (Kobayashi et al., 2018). This technique when you have large volumes of data is used to classify or group and to explain using existing knowledge (Harlow and Oswald, 2016).

The term association is a technique that allows obtaining the perception of consumers and their expectations of products, services, and events through the frequency of mention of topics or words in texts or dialogues (Gámbaro et al., 2014; Pontual et al., 2017; Rojas-Rivas et al., 2018).

The term association technique has two stages, the first is to find a list of words that are more prevalent within a set of tweets than others, once these words are found, the next stage is the contextualization or interpreting of the meaning of these words in the context. context of this difference (Thelwall, 2021a, 2021).

On the other hand, it is important to refer to sentiment analysis, that is, the ability to detect positive and negative opinions from text, making key component research using big data Sentiment analysis is the ability to detect positive and negative opinions from text, making it key component research using big data (Thelwall, 2019). Text sentiment analysis is an automated process of semantic examination of relationships and meanings of tweets (Alaei et al., 2019). The sentiment analysis uses the lexicon-based method (Medhat et al., 2014), which requires a predefined group of lexicons of feelings used to determine the polarity of a text which are the emotions in the text (Saif et al., 2016).

Data collection

The use of hashtags to collect information is very helpful because it allows concentrating the opinions of a specific sector or the dialogues on a specific topic in the community (Fatanti and Suyadnya, 2015). Twitter data allows us to identify communication patterns and dissemination of community information through hashtags (Park et al., 2016). Hashtags are commonly used in crisis communication in tourism, and during the crisis, there is generally an associated hashtag that allows users to obtain relevant information (Barbe and Pennington-Gray, 2018).

The data was collected from Twitter between March and April 2020, for which a group of frequently used hashtags about COVID-19 was identified, which are shown in Table 1. The data was collected through the Twitter API using the Mozdeh big data text analysis free software from big data (http://mozdeh.wlv.ac.uk), the extraction of tweets was carried out by filtering in the software the tweets that have in the text the hashtags that appear in Table 1. The data was extracted in different languages and globally, that is, from different countries (USA, UK, Japan, France, Spain, Netherlands, Mexico, Brazil), and companies in general and non-governmental organizations, managing to extract 3´559,255 tweets that met the condition that each tweet in its text could have one or more of the hashtags popular about COVID-19.

Table 1: Hashtags used about COVID-19 used in Twitter data collection.

As part of the processing, the data analysis followed the following steps using the Mozdeh big data text analysis software:

First, the data was cleaned by removing duplicate tweets.

Second, the term association technique was used to obtain the words associated with the term “risk” or “crisis” in the Twitter data collected using a quantitative process with Pearson's Chi-square statistical test, derived from a 2x2 contingency table used with a value critical threshold of 3,841. To reduce the risk of falsely believing that a word is significant when examining multiple Chi-square values, the method of Benjamini and Hochberg (1995) , which is a procedure that tests all words at the same time and shows all significant words or terms as results, this method controls the risk of false positives when running multiple tests. The statistical significance used is at the 0.1% level (***).

The literature shows evidence of the use of Twitter data extraction and the term association technique using the Mozdeh big data text analysis software to get the main topics of discussion in the tweets (e.g., Thelwall et al., 2019; Thelwall et al., 2021; Carvache-Franco et al., 2022a, 2022).