Yee Wen Yeo

1.0 - Introduction

In December 2019, a new strain of coronavirus which caused respiratory illnesses was discovered in Wuhan, China. The virus was spread globally and cost the lives of thousands of people. The COVID-19 pandemic has undoubtedly caused great impacts to humans around the world both in physical and mental health aspects as many businesses and jobs were affected significantly. Strict lockdown measures that forced citizens to stay indoors for long periods of time caused increased levels of stress and anxiety (McKie, 2020). Therefore, the main objective of this research is to extract sentiments regarding the general public’s mental health situation during the current COVID-19 pandemic and to ensure their wellbeing during the lockdown. 

2.0 - Literature Review

Knowledge modelling is constructed to find alternatives to the current problems faced in the healthcare industry. Its objective is to create a model that gives a computer the ability to process organizational knowledge. The knowledge model converts the selected knowledge into an appropriate format that is reusable with the purpose of sharing it or storing it (Lee, Kazz & Shankaraar, 2018). The knowledge model helps with the user to gain a thorough understanding of underlying mechanisms within the system.  Taxonomies are classification systems that allow us to characterize concepts alongside with their dependencies. Knowledge taxonomies enable knowledge to be represented in a graphical manner such that it follows closely with the organization of concepts in a field of study. They consist of an organized and sorted collection of certain words that are related to a particular domain and are mainly focused on the concepts of a certain topic. Knowledge taxonomies are constructed through a process of identifying, defining, comparing and classifying elements into groups.  

Figure 1: Bloom's Taxanomy Of Knowledge

3.0 -  Research Methodology

In today’s modern society, research through text mining in social media has become one of the top methods in obtaining open sourced data from the society. Social media has become one of the if not most, at least one of humans top priorities (Mahmud & Ramachandiran, 2018). As it is, humans ever since 1997, when the first recognizable social media site, six degrees came into existence, many people have since been stuck into these social contents. One of the many reasons why many would choose to do so is because social media has become a place where users would choose to place it as top priorities due to reasons such as meeting new people, expressing feelings, sharing opinions, and maybe research on trending products. As a result, social media has become a place full of data. This is why, as there are large insights of the following data, researchers are able to utilize text mining & analysis to extract, convert & make meaningful insights to derive meaning from these data (Hassani, Beneki, Unger, Mazinani & Yeganegi, 2020). Based on figure 2 below is a workflow regarding this research. 

Figure 2: Logic Of Work-Flow

5.0 -  Discussion

The following below shows the discussion of researched analysis & results obtained:  Based on the findings, different metrics of variations indicate that the terms mentioned above in this research can be used to depreciate both positive & negative keywords. However, an observation can be viewed which was even though each tweet can be derived from a different/similar twitter user, the term types in emotion context remain common in at least 50.59% for positive & 13.21% for negative based on figure 5’s concept mapping. Therefore, such illustration could tell us that based on the tweets, since users are commenting positively more often, it is safer to assume that most people’s mental health are alright as they are taking the pandemic gracefully. As we can see below, there is an illustration of a tree diagram that based on the term metrics, people are taking more into account watching or sharing video contents from youtube and etc based on Figure 6: TreeMap & average polarity of tweets containing respective keywords relating to what they are doing during the covid-19 pandemic.  

Figure 6: Tree Map & Average Polarity Of All Tweets Containing Keyword

4.0 -  Research Analysis And Results

The research analysis focuses on identifying the positive terms that are available in the tweets. Therefore, based on the results shown in Figure 3, the illustration of the top frequent positive emotional words relating to the tweets regarding words that are mostly used to convey their tweets to their fellow peers during the covid-19 pandemic. From the observation, we can see that the most frequent word used relates to the word +friend & +safe. This sort of illustration could possibly show a hidden message from possible tweets. Such scenarios could relate to friends in social media tweeting is he or his friend safe during this pandemic? Therefore, from this, we are able to fruitfully derive how users feel based on this pandemic thus predict if a user suffers significant mental health issues from this pandemic. Naturally, if a user uses positive terms, this would illustrate that he or she is mentally healthy. Based on Figure 3, the most linked and correlated to the term +friend among each term in the positive section are +safe, +best, +great, +thanks & +happy displayed on the concept map. This shows that all those terms & more are related to +friend in the positive term cluster. As we drill deep into the concept map, word illustration relating to +safe is linked to the category positive in emotional context such as “care”. In emotional lexicons, the category care could lead to a great sense of sympathy. Therefore, the concept map shows relationships between the two words friend & safe illustrating caring characteristic. 

In contrast, while taking into count the positive terms, the negative terms too are illustrated. Based on the data-set, we can identify that tweets of this section illustrate the least. As illustrated in the concept map, we can identify that the most frequent of the tweets relate to the term “+hard” & “+sick”. Also, as we can see through Figure 4, it indicates that all top sub-nodes of the main node are “+order”, “+house”, “+hard”, “+working” & “+game”.As a result, such illustration could mean that throughout the pandemic, based on the tweets, people significantly find that this pandemic causes difficulties. For example, illustrating if one of the tweets relate from a covid-19 patient, it fundamentally could illustrate possibilities that a patient is deemed with a lot of health difficulties, or if a family member suffers from covid-19 and it has caused a significant trauma & etc.  

Figure 4: Concept Mapping & Word Cloud For Negative Terms

As a result, positive & negative words both indicate critical word correlation sentiments as when connected in terms, both seem to have a relationship with one another. However, even such definitions are different, having the main key terms for positive & negative to relate indicates that based on tweet users, not only 1 may tweet positively, but also may decide to perform a negative tweet. For example, Friend A may tweet that he prays for the safety of his friend but also may pray for his enemy whom is sick to suffer more from the pandemic. Based on Figure 5, the “+sick” sub-node shows sub-nodes with words “house”, “order”, “game”, “working”, “hard” while the related sub-node “+friend” shows sub-nodes related with the words “happy”, “thanks”, “great”, “best” “safe”. As the tweets are least plagued with negative terms, the sub-nodes +friend & +sick can be assumed to illustrate tweets that could be derived from positive tweet users as when examining the visual aspects of the tweets, a tweeter may feel deep sadness or happiness depicting from the twitter tweets from similar twitter account users. 

Figure 5: Concept Mapping for +sick & +friend Terms

Based on Figure 6, the Treemap illustration indicates with dark metric words(positive) and light metric words(negative) & also a Tweet polarity barchart illustrating with index metrics for most common categories of tweets. Therefore, as we can identify based on metrics that users initiate to watch & share youtube videos often during this pandemic, it illustrates that the mental health derives in a more positive matter as users are performing healthier habits rather than illustrating exampled despairs such as;

• leaving home despite the pandemic for dinner with friends due to open house events
• sleeping in 24/7 without getting some form of exercise
• high social media interaction & game playing negative sentiment based games such as (GTA) Grand Theft Auto and more.

As a result, since majority of tweets illustrates tree map & average polarity concepts relating to more healthy and graceful activities, the mental & also physical health status of a user can be assumed to be illustrated for a user to be positively healthy. By reviewing Table 1 below, we can see some examples of positive & negative keyword sentences. 

6.0 - Implications

For this research, our main goal of this study is to classify, categorize word polarity vectors of twitter tweets to identify the mental health status of users based on the health activity habits that they do during covid-19 to ensure they are taking the pandemic positively or negatively. This technique illustrates natural language processing techniques as based on word polarity, they are categorized into positive or negative sentiments using emotional lexicons to identify the type of activity that might illustrate to affect their mental health. As gauged through the tweets, the results from this analysis show that it can be useful for hospitals as these data reviews would be of beneficial support for hospitals, government agencies and etc to forecast & see the mental health levels & personal psychological issues of the general public post covid-19 pandemic. These insights and knowledge can be used to allocate resources and personnel in hospitals more efficiently so that at any given point of time, there is always sufficient numbers of personnel to tend to the amount of patients who are seeking mental health treatments. For example, when the general public’s sentiments are negative and it can be seen that sadness levels are high, hospitals can allocate more staff at that period of time in anticipation of the increasing number of patients that are going to the hospitals to seek mental health psychological treatments.

Similar Research Conducted: Mental Health Psychological Issues Prior & Post Covid-19

A study was done by a team of specialists on cognitive neuroscience and psychology to survey the general public in China to understand their levels of psychological impact during the COVID-19 outbreak. Among 1210 respondents, 53.8% of rated the psychological impact of the outbreak as moderate or severe where, 16.5% reported depressive symptoms; 28.8% reported anxiety symptoms; and 8.1% reported moderate to severe stress levels. The journal also reported that one of the factors associated with these adverse mental outcomes was social isolation due to quarantine. (Wang, C et al., Immediate Psychological Responses and Associated Factors during the Initial Stage of the 2019 Coronavirus Disease (COVID-19) Epidemic among the General Population in China. 2020). Another journal has also suggested that misinformation may adversely effect mental health by causing confusion among the general public, hence causing panic and distress (Samia et al., Impact of rumors or misinformation on coronavirus disease (Covid-19) in social media. 2020). The CDC has also identified a group of individuals who may respond more strongly to the stress of this crisis, including older people, those with underlying medical comorbidities, childrens and adoloscents, front-liners and those with preexisting mental conditions. Moreover, A study to assess the magnitude of mental health outcomes among healthcare workers treating patients with Covid-19 showed that among 1257 respondents consisting of doctors, nurses and emergency responders, 50.4% of them reported depressive symptoms, 44.6% reported anxiety symptoms and 71.5% reported distress, with nurses and frontline workers being those with most severe symptoms. (Jianbo Lai et al., Factors Associated with Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019, 2020). Therefore, it is increasingly alarming to all users that mental health problems will be an issues that needs to be addressed post & prior Covid-19 pandemic.

Recommendation To Aid Mental Heath Psychological Issues Prior To Post Covid-19

As research in mental health begins to be alarming all over around the world, one recommendation to aid prior & post covid-19 is to provide users a comprehensive online resource tool (web application) dedicated to COVID-19 screening, offering psychological support and engagement to support those in need and track the spread of the disease. By answering a list of questions including symptoms, risk factors and exposure, screened patients will then be given a directive about when to seek testing or emergency care. As we know, the mental toll COVID-19 is taking on the public and front-liners is addressed, and via an app, access to support and engagement for those in need, with specific focus in depression, anxiety and stress is available. With an app such as this, users will then be matched with therapists for specific needs. Via such recommendations, users are aided to alleviate their doubts and worries by accessing curated content centered on stress management, staying active and eating and sleeping well, including webinars, blog posts and links about staying healthy and means to help users deal with the pandemic. 

7.0 -  Conclusion 

In conclusion, we have understood that Text Mining is such a technique which is deemed useful as it allows us as users to gauge a better understanding from the large quantum bites of data through text in order to achieve identifying hidden knowledge discovery. Through the usage of text mining & gauging sentiments with sentiment analysis to understand twitter tweets and identify mental health issues relating with the current pandemic. Therefore, if tweeters are deemed to be possibly having mental health issues post pandemic, this data would serve as a guide for explaining possible reasons based on the activity done by tweeters illustrating why their main source for their particular illness. According to most of the tweets that were analyzed, we have found that most of the tweets emphasized in being positive as tweets of the following processed using natural language processing techniques fell under the category that most of the social media users are perceiving to take this pandemic positively without panicking or illustrating bad health habits that may somehow affect their mental health being in the long run. The prospect of knowing that the twitter audiences show that they are worried if their friends are safe is a healthy choice through passive stress positive sentiment thinking. Another observation that can be concluded from this study is that the quality of tweets obtained from social media is determined based on popular twitter sites illustration large target audiences. In the end, as an illustration of the twitter positive, negative & neutral tweets, it can be assumed that the mental health of all users are mostly healthy. Therefore, government bodies, hospitals and etc need not worry so much if there would be psychological side effects on human mental health prior & post pandemic. However, this result is only based on users whom tweet their status. As we know, not everyone is active on twitter. Therefore, such illustrations upon a data-set might deem bias analysis and inaccurate and further study on other online platforms are required to further validate this hypothesis.  

• Brooks, S. K., Webster, R. K., Smith L. E., Woodland L., Wessely S., Greenberg N. & Rubin G. J. (2020).The psychological impact of quarantine and how to reduce it: rapid review of the evidence Retrieved from https://www.thelancet.com/action/showPdf?pii=S0140-6736%2820%2930460-8
• Dubey, A. D. (2020). Twitter Sentiment Analysis during COVID19 Outbreak. Retrieved from http://dx.doi.org/10.2139/ssrn.3572023
• Hassani, H., Beneki, C., Unger, S., Mazinani M. T. & Yeganegi, M. R. (2020). Text Mining in Big Data Analytics. DOI: 10.3390/bdcc4010001
• Kuo, L. (2020). China virus: ten cities locked down and Beijing festivities scrapped. The Guardian Retrieved from https://www.theguardian.com/world/2020/jan/23/coronavirus-panic-spreads-in-china- with-three-cities-in-lockdown
• Lai J, Ma S, Wang Y, et al. Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019. JAMA Netw Open. 2020;3(3):e203976. doi:10.1001/jamanetworkopen.2020.3976
• McKie, R. (2020). ‘I'm broken’: how anxiety and stress hit millions in UK Covid-19 lockdown. The Guardian. Retrieved from https://www.theguardian.com/global/2020/jun/21/im-broken-how-anxiety-and-stress-hit-millions-in-uk-covid-19-lockdown
• Tasnim, Samia & Hossain, Md & Mazumder, Hoimonty. (2020). Impact of rumors or misinforma- tion on coronavirus disease (COVID-19) in social media. Journal of Preventive Medicine and Public Health. 53. 10.3961/jpmph.20.094.
• Wang C, Pan R, Wan X, et al. Immediate Psychological Responses and Associated Factors during- the Initial Stage of the 2019 Coronavirus Disease (COVID-19) Epidemic among the General Population in China. Int J Environ Res Public Health. 2020;17(5):1729. Published 2020 Mar 6. doi:10.3390/ijerph17051729
• Lee, A. S. H., Kazz, N. S. M. K., & Shankaraar, N. (2018). Beyond words of popularization mining: Reviews on comic books movies. Indian Society of Education and Environment, Chennai, India. DOI: 10.17485/ijst/2018/v11i25/118570
• Mahmud, M. M. & Ramachandiran, C. R. (2018). Social Media Dependency: The Implications of Technological Communication Use Among University Students. DOI:10.1007/978-981-10-4223-2_7
• Chan, N. K. W., & Lee, A. S. H. (2017). Voice of customers: Text analysis of hotel customer Reviews (cleanliness, overall environment & value for money). Proceedings of the 2017 International Conference on Big Data Research, (2017), pp. 104-111.ACM.Crossref