Sentiments and emotions evoked by news headlines of coronavirus disease (COVID-19) outbreak

The COVID-19 pandemic, characterized by its chronic nature and lack of an immediate cure, has presented a significant challenge to global mental wellbeing. The overwhelming amount of information, termed an "infodemic" by the WHO, has contributed to widespread uncertainty and fear. Sensationalized reporting in the media further exacerbates these anxieties. This study examines the emotional impact of this infodemic by analyzing sentiments and emotions expressed in news headlines related to the COVID-19 outbreak. The researchers focused on the period from January 15, 2020, to June 3, 2020, a time of intense media coverage and global uncertainty surrounding the pandemic. Previous coronavirus outbreaks, SARS-CoV in 2003 and MERS-CoV in 2012, caused significant morbidity and mortality, and the COVID-19 outbreak, starting in December 2019, rapidly escalated into a global pandemic with devastating consequences in terms of human lives and economic impact. The study aimed to understand if the media coverage disproportionately fueled public fear and anxiety compared to the actual death rate from COVID-19 and the use of sensational language to describe the virus outbreak (e.g., 'deadly virus,' 'public health emergency'). The researchers employed sentiment analysis, a technique in natural language processing, to classify the sentiments and emotions within the news headlines, aiming to provide empirical evidence of the emotional consequences of the COVID-19 outbreak and to inform interventions aimed at improving emotional wellbeing. The interaction between the information disseminated by the media and its impact on emotional wellbeing is crucial for understanding the overall psychological impact of the pandemic.

Sentiment analysis, a subfield of natural language processing (NLP), is typically used to classify sentiments in opinions and reviews. However, its application in the medical field, particularly concerning the analysis of public perception of health crises, remains limited. Affective lexicons, which associate words with emotions and sentiments, are commonly used in sentiment analysis. These lexicons are used in lexicon-based approaches which involves calculating the orientation of a document from the semantic orientation of its words. While lexicon-based approaches are useful for categorizing texts into sentiment categories (positive, negative, neutral), they present challenges when applied to the dynamic nature of real-time web data due to rapidly evolving topics of interest. Challenges also exist in accounting for context and valence shifters in natural language (negators, amplifiers, de-amplifiers, and adversative conjunctions), which may change the overall sentiment of a sentence or phrase.

The study analyzed 141,208 news headlines from 25 top-rated English news sources, collected between January 15, 2020 and June 3, 2020 from the COVID-19 news dashboard. The data was processed using R's 'tm' package. Several text preprocessing steps were undertaken including converting headlines to lowercase, removing punctuation, stop words (e.g., "and," "or," "the"), numbers, extra whitespace, and stemming (reducing words to their root forms). This resulted in a cleaned dataset of 6,488,545 words and 31,130 unique words. A document-term matrix was created, showing the frequency of each word across all headlines. Word frequency analysis revealed 'covid,' 'lockdown,' 'case,' 'trump,' 'death,' 'test,' 'pandemic,' 'China,' 'outbreak,' and 'virus' as the most frequent terms. Sentiment analysis was performed using the R package "sentiment", incorporating valence shifters to account for their impact on sentiment polarity. The algorithm calculated an unbounded polarity score for each sentence by considering polarized words, amplifiers, de-amplifiers, negators, and adversative conjunctions. Headlines were then classified as positive (sentiment score > 0), negative (sentiment score < 1), or neutral (sentiment score = 0). The NRC Word-Emotion Lexicon was used to identify the presence and intensity of eight basic emotions (anger, fear, anticipation, trust, surprise, sadness, joy, disgust) in the headlines. Each headline's emotional profile was determined by the majority class of emotion intensities.

The analysis revealed a strong negative polarity in the news headlines. Approximately 52% of headlines expressed negative sentiments, compared to 30% positive and 18% neutral. Figure 4 and Figure 5 visually represent the distribution of polarity scores and sentiment categories across the headlines. The box plot and histogram in Figure 6 illustrate the distribution of sentiment scores, highlighting a strong skew towards negative values, ranging from -1.85 to 1.54 with an average of -0.08. Figure 7 shows the percentage frequency distribution of sentiments. The trajectory of sentiment scores over time (Figure 8) showed an overall negative trend, with negativity peaking towards the end of the study period. The analysis also identified the specific terms contributing to positive and negative sentiments, presented in Figure 9. Negative sentiments were primarily driven by words like "pandemic," "trump," "outbreak," "virus," "death," "crisis," etc., while positive sentiments stemmed from words like "positive," "care," "global," "work," "relief," etc. Figure 10 presents the frequency distribution of emotions. Fear (20%), anticipation (15%), sadness (14%), and anger (11%) were the most frequently evoked emotions, collectively accounting for about 61% of the headlines. Keywords associated with each emotion are also detailed in the paper, indicating the specific words that contributed to each emotional response. For example, 'death,' 'quarantine,' 'hospital,' etc. evoked fear; 'death,' 'isolation,' etc. evoked sadness; and 'death,' 'trump,' 'emergency,' etc. evoked surprise.

The findings demonstrate a significant negative emotional valence in COVID-19 news headlines, potentially contributing to widespread anxiety and fear among the public. The prolonged nature of the pandemic, coupled with the lack of an immediate cure, exacerbated these negative emotions. The prevalence of negative emotions, high death tolls, and prolonged fatalities are likely to contribute to chronic stress and exacerbate mental health disorders. The study's findings are consistent with previous research on the psychological impact of epidemics. For example, the Ebola outbreak also resulted in widespread anxiety, economic hardship, and social isolation. The media's role in shaping public perception is critical; inaccurate, excessive, and sensationalized reporting, as seen in the SARS outbreak, can intensify fear and anxiety. This study emphasizes the need for balanced and accurate reporting during public health crises. The negative emotional context can influence economic factors such as consumer spending, investment, and market stability.

This study reveals the significant negative emotional impact of COVID-19 news headlines. The high prevalence of negative sentiments and emotions, particularly fear, sadness, and anger, underscores the need for strategies to mitigate the psychological effects of the pandemic. Future research could explore the long-term mental health consequences of this prolonged exposure to negative news and develop targeted interventions to improve public mental wellbeing during and after such crises. It is important to consider different cultural contexts and language variations for more comprehensive analyses of public sentiment surrounding future pandemics.

The study's limitations include its reliance on headlines only, which may not fully capture the nuanced reporting of individual news articles. The geographical distribution of news sources may also affect the generalizability of findings. The study focused on English-language headlines, limiting the representation of global perspectives. Future research should consider investigating a wider range of media content and incorporating a broader geographic and linguistic scope.