Note: Public Perception of the COVID-19 Pandemic on Twitter: Sentiment Analysis and Topic Modeling Study

Note: Public Perception of the COVID-19 Pandemic on Twitter: Sentiment Analysis and Topic Modeling Study

doi: http://dx.doi.org/10.2196/21978

Gap;

• Lack of infoveillence (information watchfulness) on covid19

Objective;

• Increase public awareness of COVID-19 pandemic trends

• Uncover  meaningful themes of concern which shown by twitter (using english language to search)

Method;

• Data mining

• Total 107,990 tweets related to COVID

• December 13 - Mar 9 2020 -- 3 months

• Frequency of keywords, 

• Sentiment analysis, and 

• Topic modeling to identify and

• Explore discussion topics over time

• Natural language processing + Dirichlet allocation algorithm

• Identify most common tweet topics

• Categorized clusters

• Identify themes based on keyword analysis

Result;

• Indicate three main aspects of public awareness+concern

1. Trend of spread + symptoms of covid -- 3 stages

2. Sentimental analysis reveals negative

3. Topic modeling -- themes divide into 3

1. Covid pandemic emergency

2. How to control covid

3. Report on covid

Conclusion;

• Producing useful information about 

• Trends in discussion of covid19 on social media

• Alternative perspective to investigate covid crisis

• Advantage

• Helps departments communicate information to reduce specific public concerns about disease

---

Main text

Infectious disease

• Wuhan pneumonia -- end of 2019

• Define by WHO later as COVID-19

Social media

• Provide rich and useful info

• Text mining - extract health information from online platform

• Social media data text mining -- 

• Tracking disease

• Assess public awareness concerning health issues

• Enabling disease forecasting

• Twitter is a good source

• COVID and social media  -- hos social media helps to tackle the disease

• Issuing previous works regarding to COVID and twitter

• Early stage of perceptions

• Aid understanding

• Emotions

• Beliefs

• Thoughts

• Important for policy makers to increase situation awareness

• Thus, make suitable interventions during the pandemics

• Two research questions want to answer by using this tool

• What is the level of public awareness in terms of sentiments + emotion toward covid19?

• What are the emergent topical themes and discourses regarding covid19?

Methods

• Using API-twitter-JAVA channel

• Collect post in English that relating to covid19

• Text analysis -- identify

• Trends

• Keywords

• Themes

• Public concerns

• Sentiments

• Main objective

• Answering research questions between end of 2019 + beginning of 2020

• Important period to determine public concerns relating to early covid19 outbreaks

Data analysis -- using python and R

• Focused on frequencies of single words in corpus of  text mining structure

• Visualized frequency through word cloud

• Content analysis

• Analysing words/msg

Sentiment analysis -- NLP

• To categorize sentiments appearing in twitter msg

• Using National Research Council (NRC) sentiment lexicon -- examine expression of 10 terms

• Anger

• Anticipation

• Disgust

• Fear

• Joy

• Negative

• Positive 

• Sadness

• Surprise

• Trust

• Negative + Positive

• Remove bc of they are classification

• thus , eight left and categorized to negative and positive emotion

Topic modeling

• Unsupervised ML analysis -- using LDA

• Identifying the most common topics in tweets

• Categorize clusters + find themes based on keywords analysis

• Coherence

• Probabilistic coherence of each topic

• Coherence score --

• Indicate whether words in the same topic make sense when extracted by those topics (relativeness, I think)

• Higher score for specific number k

• More closely related

Results

Twitter trends during covid19 pandemic

Twitter trend lines of covid19 symptoms

Frequency of keywords related to covid19

Sentiment analysis on covid19

Topic modelling

Covid19 outbreak related themes