Since we are living the Covid-19 outbreak around the world, I thought, it could be interesting to analyze the public’s reaction to Taiwan’s performance during this period by analyzing tweets on Twitter.
Being Taiwanese, I would also like to share the spirit of Taiwan can help, because, in a time of isolation, we choose solidarity, we try to contribute as much as we can in order to prevent the world’s situation from getting worse.
Ready to know more? Let’s get started!
On December 31, 2019, Taiwan contacted WHO about the emergency of viral pneumonia, in the meantime, Wuhan City also announced the discovery of the new virus. People were afraid that Taiwan may be seriously affected being only 130 kilometres away from China. On the contrary, the Taiwanese government took immediate action to set up a lot of strict policies on travelling in order to prevent the virus from entering or spreading in Taiwan.
Taiwan has only 429 confirmed cases and 6 death. This, in the four months, that the virus has been spreading like wildfire in most places.
There are several news and reports lauding Taiwan as a role model for successfully fighting the Covid-19. I, therefore, chose to investigate more on the public reaction towards Taiwan’s performance.
In order to know the public reaction to Taiwan, I used two keywords to find out the data. The first one is “Taiwan”, the second one is “Taiwan can help”. Using these two words to discover the popularity trend of Taiwan on different topics and regions. I found that the topic that people are searching is mainly related to Covid-19 in Taiwan and the relationship with the World Health Organisation (WHO). Below are the Google trend search & Buzzsumo from the period of March 1 to April 28 around the globe.
Google trend search of the keyword “Taiwan” during the period of March 1 to April 28. (100 means very popular, 0 means no data available)
Google trend search of the keyword “Taiwan can help” during the period of March 1 to April 28.(100 means very popular, 0 means no data available)
Top five articles which have the highest engagements of the keyword “Taiwan” on social media during the period of March 1 to April 28. Source: Buzzsumo
Here are also some charts based on the downloaded tweets from Twitter using the tool R. I also search two keywords, “Taiwan” & “Taiwan can help” and got 17702 and 1095 tweets respectively from the period of April 18 to April 26. Since there are not so many available tweets on “Taiwan can help”,
I mainly focus on the analysis of keyword “Taiwan” and I also combined some useful data or graph of the keyword “Taiwan can help” as supplements. We can see from the graph that there are a lot of people on Twitter talking about Taiwan during this period.
This is the scatter-plot of the keyword “Taiwan can help”. (April 20 to 26)
This is the histogram of the keyword “Taiwan”. (April 24 to 26)
This is the scatter-plot of the keyword “Taiwan”. (April 24 to 26)
The potential reach of the tweets matching “Taiwan”. (April 24 to 26)
What did I think before collecting the data?
Taiwan, an island located in the centre of eastern Asia, with 23 million habitants, is working hard to combat the virus.
Since we have gone through the SARS epidemic in 2003, we know how hard it is, and we gained a lot of valuable experience from it. That is the reason we try to help by contributing as much as we can with open, transparent data and information or medical supplies to the world.
I expect that Taiwan’s effort will be recognized internationally and get more chances to cooperate with other countries in order to protect people everywhere. Taiwan is small, but together we are stronger!
Analysis and infographics
In the analysis part, I divided into four different parts based on the 17702 tweets (keyword “Taiwan”) downloaded from Twitter using the tool R (selected language is English). The first one is network analysis (where do the users come from), the second one is topic modelling (divide the data into different topics), the third one is subject analysis (regarded to word frequency analysis), and the last one is sentiment analysis (negative or positive word in the text).
1. Network Analysis
In the network analysis, I would like to know where do the users come from and what is their reaction in order to understand the popularity trend of “Taiwan” around the world. When I calculated the users by countries, I found that there were some data not available in the country column, therefore, I grouped the location by country to show the number of users on the map. We can see that except Taiwan itself (1127 users), there are a lot of people in the US (1109 users) are talking about Taiwan, and the rest of countries are mainly around Asia (the results may be restricted to some certain extent since the selected language is English).
The users around the world on Twitter talking about “Taiwan” from April 23 to 26.
This graph shows the top ten highest reactions to the keyword “Taiwan” based on countries (except Taiwan). The reaction score was calculated based on four different elements, which were the sum of retweets, favourite count, reply count and quote count. We can tell from the graph that the US has the highest reaction, which means that the people in the US are highly engaged in this topic.
The reaction score between different countries. (April 23 to 26)
2. Topic Modeling
Topic modelling is a method to find a mixture of words which is associated with each topic, here I divided into 6 different topics.
Here are the 8 most frequent words in these 6 models.
The scatter-plot shows the retweets in each topic (April 24 to 26). (*the group called “NA” is because some tweets are too short to assign into one topic).
This graph shows the tweets, reactions and the reach in each topic (April 23 to 26). ‘Tweets’ is the count of the number of tweets in each topic, the reaction is calculated based on the sum of retweets, favourite count, reply count and quote count, and the reach is referred to the followers count. We can discover in this graph that it has a similar trend and shape as the above one.
Since topic two has the lowest score, therefore, I would like to dig into the reason behind it. I found that it is mostly affected by the count of followers and the account creation time. (please see below these two graphs for the reference)
3. Subject Analysis
In the subject analysis, I would like to know the word frequency in the text. In other words, we can discover the words that appeared a lot of times on Twitter, and we can predict the trend or the hot topic among the users on social media. For example, Taiwan started to donate the medical supplies on April 1 to the countries in need, and we even did the crowdfunding project on April 10, in order to publish the news on the New York Times to share the spirit of Taiwan can help. Therefore, there might be some users started to discuss this topic on Twitter. Let’s see which words are popular on Twitter!
We can tell from the word cloud and the graph that the most frequent words are mainly related to the Covid-19 and some Asian countries names are mentioned as well.
The reason probably lies in the fact that people are comparing the situations in each country around Asia, and also Taiwan has a closed relation or maybe some medical cooperation among these countries. Besides, we can also observe that the words are generally positive, which means that Taiwan shaped a good image on these users. We will explain more details in the next part—sentiment analysis.
4. Sentiment Analysis
Sentiment analysis is the interpretation and classification of emotions (positive, negative and neutral) within the text data. There is a criterion to determine the sentiment analysis score, -1 means very negative, 0 means neutral and 1 means very positive. Therefore, in this part, I would like to identify users’ sentiments based on their online conversations and feedback toward the topic related to Taiwan.
We can see from the above two graphs that the average sentiment score is 0.03386449, it can be described as neutral to slightly positive, but there is still a big room to improve in order to be regarded as very positive. So, let’s now move on to what exactly are these positive or negative words that affected the result of the score.
From the observation of the above two graphs (positive and negative words), we can discover that the count of number in the positive words is much more than the negative ones.
As a result, we can indicate that the whole text data set is considered to be positive, which means that most of the users they gave affirmative feedback on Taiwan’s performance during the outbreak of Covid-19.
What did I find? Was it different from expected?
In conclusion, I found that the result meets 80% of my expectation, but there are two things that can be improved.
First, we can see that most of the highly engaged users are still from Taiwan instead of other countries, therefore, Taiwan still needs to put the effort in building the international image and reputation, for the sake of letting the world knows that although we are small, we have the ability to help when the globe went through this severe pandemic.
Secondly, I expected that the sentiment analysis score would be a little bit higher than I thought. I think it is probably because there are some frequent negative words used to describe the Covid-19 situation in the text (for example virus, pandemic and death etc.) which affected the final average score.
All in all, I think the public generally recognized Taiwan’s outstanding performance in fighting against the Covid-19 virus and we will continue to cooperate with the world to show our responsibility as a member of global citizens.
Charlotte Kuai is a Taiwanese Student pursuing her Master of Science in Big Data at TBS, Toulouse. You can follow her Facebook page ‘Letscourage’ or write to her (email@example.com).