Keywords

1 Introduction

Since the second half of the year 2019, COVID-19 has been sorely hitting the world. The seriousness of this pandemic can be determined by the uninterruptedly growing COVID-19 cases around the world. As per the current situation, countries like the USA, Russia, Brazil, and India have been severely affected by this virus [1]. Several lockdowns were imposed worldwide in different countries to minimize the outspread of the disease. Due to these unprecedented situations in the lives of all human beings alive today, the behavior and emotions of human beings have been greatly affected [2]. As one could expect at this time of total lockdown, it is but natural for a person to have different sorts of emotional swings. A lot was happening within a fraction of days, that there were mixed emotions [3]. The physical movement of people came to a halt to maintain social distancing. The sudden outbreak of the disease made the authorities put in a tough situation to handle the medical facilities for treating the deadly virus [4] as well as the psychological disorders that were caused by the imposed lockdown. Understanding the emotional and mental wellbeing of the citizens of the country plays a major role in sustaining these kinds of pandemic situations. The World Health Organisation (WHO) [5] also issued several guidelines to maintain the healthy mental state of people throughout the pandemic. While some people were dealing with the emotional trauma, fighting with the negative psychological effects while other people were finding ways to entertain themselves and getting more and more creative in these times of home confinement [6, 7]. To assist numerous administrations and organizations, it is extremely necessary to comprehend the emotional state of people of the country so that they can be helped in overcoming the psychological imbalance that has occurred throughout this time [8]. Moreover, technology has grown by leaps and bounds in today’s modern world [9]. The whole tedious process of writing and publishing news articles has been simplified using social media [10]. Immense amount of data is produced every day every minute [11]. Nowadays, online social media websites are frequently used as a source of analyzing and investigating the emotional state of people. Social media platforms play a vital role in the time of trouble to evaluate the mental and emotional health of people around the world [12]. Particularly, social media platform Twitter provides a handy mechanism to express one’s viewpoint in short and crisp messages. The challenge is to identify, process, and combine various data sources to reach intelligent decisions [13].

When it comes to highly populousFootnote 1 countries like India, COVID-19 cases have significantly escalated. Owing to the heterogeneous demographics of India, it took just three months for the number of cases to reach almost thousand times the cases that were at the commencement of the spread. Furthermore, living in the largest democracy [14], people are free to express their feelings and thoughts on all the incidents happening around them. Every individual has the right to convey their opinion on political, social, economic, and cultural matters.

Taking into account the seriousness of the pandemic, it is very essential to persuade the public to reduction of the spread of COVID-19 and adoption of sustainable measures. Several initiatives were taken, not only by the Government of India but also by the citizens. The government has set up helpline centers to help people (telephonically) with emotional issues during the pandemic. The adoption of these sustainable measures helps to mitigate such crisis situation and make the public future-ready. During COVID-19 also, people took healthy participation in spreading awareness about the disease not only in cities but also in village areas. In order to minimize the outbreak of the disease, the Indian government gave several instructions to people from wearing masks to maintaining social distancing among each other. Table 1 shows the major events that caused a havoc of emotions amongst the people. These events played a key-role in triggering emotional outbursts on Twitter.

Table 1 Major events that took place in India during the lockdown period
Full size table

In this chapter, we present an internet portal visualizing the eight-pointer emotional spread amongst the people of different states of India during different phases of lockdown. Our application does real-time monitoring of tweets and displays the tweets of people related to various hashtags as mentioned. The internet portal allows to select the state as well as particular lockdown for which the emotional spread is to be visualized. The analysis conducted considers textual information contained in the tweets while exploiting the linguistic features. Subsequently, emotion lexicon (NRC Word-Emotion Association LexiconFootnote 2 [15]) has been utilized to understand state-wise and lockdown-wise emotion. This emotion analysis has been conducted on the tweets posted by people during different lockdown stages. In this analysis, information has been extracted from Twitter using Twitter API.Footnote 3 The tweets associated with hashtags #CoronaVirus, #LockdownDiaries, #Lockdown, #Covid-19 that were posted on Twitter in different states of India in the duration March 2020–June 2020, have been analyzed and categorized into one of the eight emotion categories—Anger, Anticipation, Disgust, Fear, Joy, Sadness, Surprise and Trust [16].

The rest of the chapter is organized in the following manner: Sect. 2 describes the related work completed in the field of emotion analysis during COVID-19 attack. Section 3 describes the proposed methodology and gives insight into the approach used to analyze the tweets of the dataset. Section 4 presents the results obtained through the analysis and a brief discussion. Section 5 concludes the chapter and throws light on the future scope of the project.

2 Related Works

Understanding the emotional state of people in such difficult times of crisis, not just pushes the public authorities to survey old guidelines, yet in addition, helps in forming new rules and undertake measures that can propel the masses and re-establish their physical and emotional wellbeing. Consequently, to examine the public feelings on different occasions during this pandemic, a few investigations have been led. There are a few models proposed for dissecting the public feeling during COVID-19 pandemic.

In [17], the proposed approach examines the emotions of Weibo users in China—gender wise and age-wise. The outcomes additionally showed that the expressions of concern like “Wellbeing”, “Family” and “Demise” expanded fundamentally demonstrating that with sitting back, individuals were getting more worried about their family and wellbeing. Another work [18] examines the mental health of individuals during lockdown stage 2 and lockdown stage 3 on Twitter dataset and investigated the assessment on web-based businesses (e-commerce) during this pandemic. There was a sure plunge in the level of joy, fear, and trust in lockdown stage 3 when contrasted with that of lockdown 2 while there was a sure ascent in level of disgust, anger, and anticipation. Analysts have likewise broken down the news-headline features of Coronavirus and performed sentiment and emotion association [19].

Cao et al. [20] contemplates the effect of Coronavirus on students in China. The study uncovered that approximately 25% of students have encountered tension in view of this COVID-19 episode. The outcomes inferred that the danger factor and the postponements in scholastics were the fundamental explanations behind expanding tension while factors like living with family and having consistent family earnings were defensive variables against experienced nervousness during the COVID-19 flare-up.

Landicho-Pastor [21] discussed the students’ sentiment on online schooling strategy dependent on an open-ended poll. Their investigation uncovered that many students were not ready for the online method of training and were stressed over the elements like the network issues at their place.

As the pandemic began, numerous analysts were interested to gain understanding of major worries of Twiterrati on COVID-19. Abd-Alrazaq et al. [22] examined 2.8 million tweets found out twelve concepts. Heffner et al. [23] considered the public readiness to self-isolate by breaking down the conclusions on two kinds of self-isolation rules, either undermining or written in convincing language. Their outcomes showed that despite the fact that individuals evoked negative supposition for government-imposed rules, they showed readiness toward social distancing.

3 Data and Methodology

The methodology used in this chapter includes a set of major tasks performed to achieve the desired visualizations. Figure 1 shows the methodology used in the analysis.

Fig. 1
figure 1

Methodology

Full size image

3.1 Dataset

The dataset has been curated from Twitter using public streaming API (see footnote 3) on India—specific tweets to perform analysis. The Twitter API facilitates fetching of real-time data by providing certain hashtags and specifying locations. The hashtags used for this analysis are #CoronaVirus, #LockdownDiaries, #Lockdown, and #Covid-19. The final dataset is taken into account for analysis, mainly consisted of tweets, date, time, and their locations from where the tweets were posted. The considered duration is 01-03-2020 to 09-06-2020. The word—emotion association was performed through NRC emotion lexicon (see footnote 2) which consist of scores for eight emotions: anger, fear, anticipation, trust, surprise, sadness, joy, and disgust [15]. The dataset was cleaned by removing the tuples of incorrect data and ‘Not a number’ values. The tweets were pre-processed using Natural Language Processing techniques [24]. Application of machine learning to explore the Impact of Air Quality on the COVID-19 Fatalities is also studied [25, 26].

The text of tweets in the form of a string was first word tokenized into individual words. Then, as a step towards cleaning of tweets—slang, misspelled words, hashtags, URLs, and emoticons have been removed [27]. Additionally, the tweets have been converted to lowercase. User mentions along with re-tweets were removed, followed by elimination of punctuations, special symbol characters (except a–z, A–Z). A list of English stop words has been imported from NLTK library which is used to remove the stop words present in tweets.

3.2 Two-Way Emotion Characterization

In this chapter, a two-way emotion characterization of tweets has been presented state-wise and lockdown-wise:

  1. (a)

    State-Wise Emotion Analysis—In this analysis, tweets of the people from different states of India have been collected, cleaned, and analyzed. The approach processes each tweet and categorizes it into one of the eight emotions. The results of this analysis are visualized on the map of India, with each state depicting the intensity of a particular emotion.

  2. (b)

    Lockdown-Wise Emotion Analysis—In this analysis, tweets posted during four lockdown phases observed in India have been taken and dissociated according to different states. Tweets have been cleaned, analyzed, and categorized into one of the eight emotions. The results were analyzed using stacked bar charts for different states of India.

3.2.1 State-Wise Analysis

The dataset is dissociated according to different states. The tweets are POS tagged to help identify adjectives and adverbs. The state-wise emotions are given a score based on the most frequent words found in tweets and NRC emotion lexicon list. A ‘state-wise-emotion-count-dictionary’ dictionary is formed which contains words with their occurrence frequency. This dictionary is used to calculate the sum total of eight emotions. With these scores, the heat maps of India have been created depicting respective state-wise emotion intensity levels (ref. Sect. 4, Fig. 3). The heat maps of India for different emotions have been developed using boundary shape file “Indian states.shp”. The shape file included features: state name and geometry. The merged data frame was formed by combining state emotion count with the Indian state shape file (.shp file). The plots of separate emotions were created by using Matplotlib’s subplot and plot function in Seaborn style.

3.2.2 Lockdown-Wise Analysis

The lockdown-wise analysis is carried out with the help of dates extracted from tweets in the dataset. The same has been presented through stacked bar charts (ref. Sect. 4, Fig. 4). This lockdown-wise analysis has been carried out for lockdown 1, lockdown 2, lockdown 3, and lockdown 4 during COVID-19 pandemic. The lockdown phases are listed in Table 2.

Table 2 Lockdown phases
Full size table

Table 2 depicts the dates of lockdown issued by the Indian government. The lockdown-wise emotion analysis was conducted for separate states of India. A word count dictionary was created containing words with frequency segregated according to four lockdown phases. This was then categorized into different emotions based on the NRC emotion lexicon list. Stacked bar plots were obtained for top 12 states and are depicted in Sect. 4, Fig. 4, using Matplotlib.

3.3 Internet Portal

An internet portal has also been launched as an outcome of this analysis, as shown in Fig. 2. It depicts graphical figures that quantify public emotions during COVID-19. The portal provides the users with two menus—based on the states and the emotions.

Fig. 2
figure 2

Internet portal. https://emotiontrackerindia.herokuapp.com/

Full size image

4 Results and Discussion

The pandemic caused by coronavirus has been one of the most unexpected and grave casualties for the whole world. This section presents visual analysis depicting a tremendous downpour of public emotions expressed towards the lockdown on an eight-pointer scale of emotions: anger, fear, joy, disgust, anticipation, surprise, sadness, and trust. Figure 3 presents the state-wise emotion wise analysis.

Fig. 3
figure 3figure 3figure 3figure 3

a–h Country-wide heat map of all emotions

Full size image

Despite massive amounts of efforts by the government, Maharashtra, Delhi, Haryana, Uttar Pradesh, Karnataka, Madhya Pradesh, Tamil Nadu, and Rajasthan have been worst affected. Figure 3 depicts this analysis. There is an apparent relationship between the states where more cases were reported, and huge mix of emotions were also reported from those states only.

Some states like Goa and Manipur have managed the pandemic situation commendable as depicted in Table 3. People have been quite enthusiastic about the lockdown time. They have considered it a time for personal productivity and growth. Some are also happy spending time with their families. While others who have lost their jobs have been really worried and upset. Economy of the nation has been completely shattered by the situation. This has been depicted in Sect. 1, Table 1.

Table 3 Reported, recovered, and deceased cases of worst hit Indian states
Full size table

Being the capital of the nation, Delhi has had huge responsibilities of catering to everyone’s healthcare needs. Many people have been moving to Delhi for their treatment as it is considered to have the best medical facilities. This has been a reason for chaos and worry amongst the people and has caused a shortage of hospital beds and other medical facilities like ventilators and oxygen cylinders.

Another issue that has been an area of concern throughout was public gatherings like Tablighi Jamaat and mismanaged movement of people who lost their jobs across state borders. There was a shortage of food and a lack of facilities for these laborers to go back home. These people had to walk back home covering hundreds of kilometers all by themselves. This has been the major cause of anger throughout the nation as these people were highly susceptible to catching infection. During the lockdown phases, international travel had completely stopped. This was another cause of stress for the Indian citizens stuck outside and for their families. The government played a very positive role in resolving this issue by trying to get all the people back to their country. This gave people relief.

Figure 4 and Table 4 depict lockdown-wise emotion analysis of states. In the case of Maharashtra, it can be clearly observed that there was a sharp rise in anticipation and sadness amongst the citizens. This has been the general pattern in almost all the states like Delhi, Gujarat, West Bengal, Rajasthan, Punjab, Goa, and Karnataka. These states have seen massive growth in Coronavirus cases even after following the strict rules of the first lockdown. These emotions tend to lower down during lockdowns 3 and 4 which may be due to the increase in recovery rates. On the other hand, cases in Uttar Pradesh reduced in the second lockdown, and hence the counts somewhat decreased there. The same decreasing pattern was observed for the 3rd and 4th lockdowns. The number of Corona cases in Madhya Pradesh was not very high during lockdowns 1 and 2 but the people there were shocked by the growth during lockdowns 3 and 4. This led to an emotional upheaval amongst the people. Observing the general trend, it can be quoted that lockdown 2 had shown the maximum rise in the number of COVID-19 cases which decreased down during lockdown phases 3 and 4. The emotional pattern of the people was widely linked to the number of reported cases, recovered cases, and deceased cases during different lockdown stages.

Fig. 4
figure 4figure 4figure 4figure 4figure 4figure 4

a–l Lockdown-wise analysis of emotions in 12 states

Full size image
Table 4 Emotion depicted by Indian states during different lockdown stages
Full size table

Overall, pandemic time has been stressful as expected. But at the same time it has brought the people of India together, fighting for one mission—to keep the nation safe and healthy. India has again proved that there is unity in diversity. People of India, under all circumstances, stand strong together.

5 Reflections

This chapter studies the social consequences of COVID-19. The pandemic posed unusual challenges to human behavior and mental health [27]. In this study, tweets posted across India during different lockdown phases have been analyzed to gain an understanding of the psychological health of the people. Heat maps have been plotted on the map of India, denoting the intensity of emotion in different states. This kind of analysis can help policy makers to conduct psychosomatic assessments of people in order to provide support in times of crisis. The notion of sustainability somewhere lies in settling the recent concerns and also parallelly focusing on long haul answers for battling with comparative issues in future. With everything taken into account, lockdown would not have been forced in the event that we would have been ready for such pandemic circumstances physically, intellectually, emotionally, and financially. Undertaking right sustainability measures at the perfect time fortifies the system alltogether. These proactive measures likewise search for long haul help habitats for individuals worldwide. We should recognize the issues deeply and afterward work on them. A nation leads by its residents, and it is similarly imperative to keep up sanity and psychological wellbeing of individuals. This chapter also puts forward an internet portal, as a by-product that can show intensity levels of different emotions in different states of India alongwith different timelines.

The scope of this chapter is not only limited to the tweets on Twitter. The analysis can be broadened if we move to other social media platforms such as Facebook, Mastodon, Gab, and Peeks. The work presented in this chapter can be extended to aspect-based emotion analysis on the dataset. Furthermore, this analysis can be scaled to encompass different countries with different native languages.