OPINION MINING CHALLENGES AND CASE STUDY: USING TWITTER FOR SENTIMENT ANALYSIS TOWARDS PFIZER/BIONTECH, MODERNA, ASTRAZENECA/OXFORD, AND SPUTNIK COVID-19 VACCINES.
Abstract
The increasing number of users on the internet and online servers such as Twitter,
Amazon, Yelp, and Facebook has significantly motivated people not only to use the
internet for their transactions, but more importantly to voice their opinions about servers,
products, policies, etc. Sentiment analysis is the task of classifying opinions about
specific topics, such as servers, products, and policies into positive, negative, and neutral
categories. The field of sentiment analysis was introduced about 20 years ago, and it has
widespread applications and models in different domains such as marketing, risk
management, and politics. Moreover, opinion mining has significant impacts on
businesses, servers, politics, and other significant strands of society, making it important
to areas that benefit several fields in real life. This research introduces various problems
facing the opinion mining field relative to the Covid-19 vaccines and suggests the
parameters of the application to illuminate the definition of the problem and its effect on
the opinion mining field and application. Those problems include sentiment analysis
accuracy, sentiment lexicon, natural language processing issues, fake opinions,
subjectivity detection, and opinion summarization. Each problem contains an important
variable in the content of the text that can have an important role in an opinion mining
task model. The field of sentiment analysis has made major leaps in its application and
field. However, this research presents shortcomings in the field of sentiment analysis
regardless of the importance of the key aspects of this field. Included as well is a case
study related to the sentiment analysis task using Twitter for sentiment analysis towards
Pfizer/BioNTech, Moderna, AstraZeneca/Oxford, and Sputnik Covid-19 vaccines.
Applying sentiment analysis can be done in many languages source codes such as C,
Java, Python, etc. This research applies sentiment analysis using Python language source
code, since it has a library that supports data analysis. The case study will include preprocessing, sentiment analysis, and visualization. The aim of this thesis is to introduce
and illuminate the challenges facing the opinion mining (sentiment analysis) field.