University of Nebraska - Lincoln University of Nebraska - Lincoln
DigitalCommons@University of Nebraska - Lincoln DigitalCommons@University of Nebraska - Lincoln
Library Philosophy and Practice (e-journal) Libraries at University of Nebraska-Lincoln
10-22-2020
Tapping Twitter Data for Analyzing and Visualizing Public Tapping Twitter Data for Analyzing and Visualizing Public
Sentiments on Censorship Sentiments on Censorship
Naveen Kumar Yadav
CFEES, Defence Research Development Organisation, New Delhi, India.
Akhilesh K.S. Yadav
Centre for Library and Information Management Studies (CLIMS), Tata Institute of Social Sciences,
Mumbai, Maharashtra, India.
Follow this and additional works at: https://digitalcommons.unl.edu/libphilprac
Part of the Data Science Commons, and the Library and Information Science Commons
Yadav, Naveen Kumar and Yadav, Akhilesh K.S., "Tapping Twitter Data for Analyzing and Visualizing Public
Sentiments on Censorship" (2020).
Library Philosophy and Practice (e-journal)
. 4468.
https://digitalcommons.unl.edu/libphilprac/4468
Tapping Twitter Data for Analyzing and Visualizing Public Sentiments on Censorship
Naveen Kumar Yadav
1
& Akhilesh K.S. Yadav
2
1
Junior Research Fellow, CFEES, Defence Research Development Organisation, New Delhi,
India. Email: [email protected]
2
Assistant Professor, Centre for Library and Information Management Studies (CLIMS), Tata
Institute of Social Sciences, Mumbai, Maharashtra, India.
Email: [email protected]
Abstract
The main objective of this research study is to analyse and visualize Twitter data with tags
“#Censorship”. A connection was established with twitter using Twitter API, and receiving the
tweets on Google Spreadsheets. Data visualization was performed using various tools such as
Voyant Tools, Tableau, Google Spreadsheet and Orange in order to generate different
visualizations based upon, language, geographical areas, retweets etc. The sentiment analysis was
performed for the sentiments that were attached to the given set of data by the public in their
respective tweets. The 23680 tweets were retrieved during the data collection time and there were
13,771 retweets out of these tweets. The most popular application for using twitter by the users
was Twitter Web Client which constituted of the 33.67% (7972 tweets); the second most popular
app was Android with 23.61% (5592 Tweets) and Twitter for iPhone stayed at third place with a
share of 20.07% (4753 Tweets). The most frequent used Hashtags (#) in the tweets were #Twitter,
#Facebook, #Google, #YouTube etc. Results show that negative tweets are enormously higher
than the neutral sentiments.
Keywords: Censorship, sentiment analysis, public sentiments, text mining, data visualization,
text visualization
Introduction
Social Media is not a strange topic in the present-day environment. Social Media has been a
frequently used term in the recent past on various issues around the world, where lives of common
people are mingled with the computer and internet. The picture that comes into our mind when we
hear the term social media is of different applications such as Facebook, Twitter, WhatsApp,
YouTube etc. Basically, the word describes the interaction among people or groups of people
where they are able to create, exchange or share information and ideas in the web environment
(Ahlqvist, Back, Halonen & Heinonen, 2008). The new generation media and networking has
given rise to the Social Media. The most widely used social media applications includes
WhatsApp, Facebook, Twitter, Instagram, YouTube etc. which are both computer and
mobile/tablet-based. People who are well versed with them can access these sites through Internet
at home, school, offices etc.
The computational treatment of opinion, sentiment and subjectivity through text is known as
Sentiment Analysis. (Pang and Lee, 2008). Kharde and Sonawane (2016) defined Sentiment
Analysis as, “the process of the mining of attitudes, opinions, views and emotions automatically
from text, speech, tweets and database sources using the techniques of natural language
processing”. It includes classifying of someone’s opinions present within the text into different
sentiments such as, “positive”, “negative” or “neutral”.
It is often regarded as analysis of subjectivity and also opinion mining. There are many tasks
performed under the name of sentiment analysis which includes sentiment extraction, sentiment
classification, summarization of opinions etc. Sentiment Analysis mainly focusses in analyzing the
sentiments, attitudes, opinions and emotions of people towards different products, individuals,
topics, organizations and services.
Social Media Platforms, and especially microblogging in the recent years have enabled a sudden
rise of unstructured content including opinions with regard to a variety of topics and have provided
a cheap and valuable source for analysis of opinions about different organizations and individuals.
It is not feasible to analyze all the data without taking assistance of automated computation. For
an average person, it is very challenging to identify and summarize important information in huge
amount of data (Wani & Jabin, 2018), along with this, the amount of data created presently over
Social Media everyday makes it an impractical task for human processing (Bello-Orgaz, Jung &
Camacho, 2016).
Sentiment Analysis may be performed at diverse level such as document, sentence and aspect
levels. The most common approach is document level classification. It classifies the whole
document by assuming that there is single entity opinion expressed in the whole document. For
instance, customer reviews (Zhang & Liu, 2017). However, it becomes harsh for some documents
which might include opinions with regards to diverse entities or even the applications may require
classification regarding various aspects of the entity.
A lot of work has been done in the field of “Sentiment Analysis on Twitter” in the past decade or
so by researchers around the globe. Park, Ko, Kim, Liu & Song (2011) and Bakliwal, et al. (2012)
tried to experiment opinion mining using news stories or tweets by users respectively to identify
political views and opinions. In the approach towards twitter, it was also considered if the tweet
contained sarcastic remarks or not.
Tweets are short texts of length of maximum 280 characters. Every day, millions of tweets are
generated on different kind of issues around the world to share and discuss the thoughts and views
of the users. Sentiment analysis on Twitter has been used for several applications. Mostafa (2013)
in a recent work identified 4 major areas in which sentiment analysis was used which includes
“political orientation extraction, stock market predictions, product reviews and movie reviews”.
The terms, ‘sentiment analysis’ and ‘opinion mining’ are used identically, where the later one is
used in the area of industries. The extent of Sentiment Analysis is to ascertain the views, emotions,
judgement or attitude of a person towards a particular topic of concern. The biggest problem that
researchers face is that almost every document that they want to harvest is written in Natural
Language Text, which is often delineated as unstructured data. Other problem that rely is the
complication to distinguish between the nuanced usages of words and it also becomes extremely
problematic to detect sarcasm or noisy data. Most of the approaches that are taken in order to
classify a document after properly appropriating some positive/negative classifiers and then collate
their results. However, in this case, Sentiment Analysis is a categorization problem in which the
documents are categorized into two categories, positive and negative.
Objectives of the study
1) To understand the public discussions on censorship on Twitter microblogging.
2) To identify the locations, trends and applications used by public for frequent posts on Twitter.
3) To apply text mining techniques to analyze unstructured text content on the topic of censorship.
4) To find out the pattern and context with censorship.
Research Methods
The main motive of this phase of the research work was mainly to set up the webserver and test
the Twitter API to pull in specific tweets in real time using Google Spreadsheets. A spreadsheet
with tweets was created from January 5, 2018 to February 9, 2018. The tweets with the hashtag,
“#Censorship”, that were later used for data analysis and visualization. A number of tools were
used in order to achieve the desired results, as mentioned below: a) Twitter microblog is used to
comment, respond and amplify on real time events. Microblogging became very popular since
Twitter and Jaiku came into existence. b) Voyant Tools: Voyant Tools is a web-based software
developed for text reading and text analysis. c) NodeXL: It is an application for network analysis
on general purpose and supports network overview, discovery and exploration. d) Google
Spreadsheets: Google spreadsheet is a web-based application that allows users to create, renovate
and alter spreadsheets and share the data live online. The Ajax-based program can work both with
Microsoft Excel and CSV (comma-separated values) file formats.
Data analysis
The total number of tweets archived during the data collection time was 23,680 and there were
13,771 Retweets. The amount of unique archived tweets was 23,081 which is used to monitor the
quality of the archived data. The first tweet was retrieved on 5th January’2018 and the last recorded
one was on 9th February’2018. The average tweet rate count was approximately 0.1 Tweets per
minute (this is calculated after each time archive has been run). There were around 1819 IDs which
were mentioned in replies.
The top Tweeterer in the entire data collection period who posted with maximum number of tweets
had the username “schestowitz” and posted a total of 234 tweets. The user received mentions in
70 tweets and got 5% of the total retweets. The second in the list with 174 tweets was username
“kosmofilo”. The third place was taken by username “AnonymousVideo” with 96 tweets, 19
mentions and only 1% of retweets. Username “DailyBrian” posted 88 related to censorship and
received 352 mentions and 16% Retweets. The fifth top username “BrianBrownNet” with 86
Tweets, 19 mentions and the highest number of Retweets with 99% Retweets.
The figure-1 shows the number of tweets that were recorded by the server on each day. It can be
seen that there has been some un-common rise in Tweets based on the Tags ‘Censorship’ on 10th
and 11th of January when the number of counts raised up till 2163 in one single day which indicates
that there could have been some issue/news event occurred related to ‘Censorship’ at that particular
time. The issue that must have achieved so many tweets was the announcement of the controversial
Bollywood movie ‘Padmavat’. The least number of tweets were recorded on 8
th
January with just
320 tweets.
Figure 1: Twitter trend on censorship
The most used application for using twitter by the users was Twitter Web Client which constituted
of the 33.67% (7972 tweets) of the total tweets posted. The second most popular app was Android
with 23.61% (5592 Tweets) and Twitter for iPhone stayed at third place with a share of 20.07%
(4753 Tweets). The other top application in the later order included Twitter for iPad, Twitter Lite,
TweetDeck, Hootsuite etc. There were around 243 different number of applications that were used
by the users while accessing their twitter accounts and posting tweets.
Figure 2: Applications used for tweets
The figure 3 shows the count of the most frequently used language by the users while posting
tweets. As observed, the most widely used language was English in the Big Blue Bubble with a
total of 21,497 tweets posted alone. Most of people use English as a language to post the content
so that others are able to understand, with English being the most widely used language in the
world. In order to reach the global audience, users on twitter use English language while posting
their content. The second most widely used language turned out to be British English (en-gb). We
know that there is a difference between the American English and the British English and the
British English, mostly spoken/used by the people belonging to United Kingdom has been the
second most used language for posting content. There was a little difference between the languages
used on 3
rd
and 4
th
place. The other languages were the European Languages German (de) and
French (Fr). The fifth language that was used for posting content by users was Spanish (es). The
figure shows that European Languages somehow dominate the Twitter usage count after English.
Figure 3: Bubble Graph of the Languages used in Tweets
It was found that the most of the people who tweets on #censorship are from various countries e.g.
India, Australia, New Zealand, South Africa etc. The least posting countries turned out to be the
South American Countries of Brazil, Paraguay, Argentina along with Canada, Some of the Middle
East Nations and South East Asian Countries. Major technologically advanced countries such as
United States of America and China stayed on the second to last spot in the list with an average of
2000 tweet counts.
Figure 4: User’s Location
It depends on the users if they wish to allow access to their location or not on Twitter. Sometimes,
this data is not clear, as it cannot be justified with many users not letting the applications to access
their location. However, it can be articulated by cleaning up the data and taking out the separate
data with location coordinated in it. By using Google Maps feature, the region-based information
can be achieved for doing research on the location of the users. It can be observed in the figure
that the most located regions by the locater is United States of America, United Kingdom and
India. Along with this, there are few locations traced from Australia, Africa and South America as
well. The least recorded regions turned out to be China and Russia with restrictions imposed on
the users by the Government authorities for not allowing location access to the applications or may
be because of users are not allowed to post anything related to censorship.
Retweets is a re-posting of a tweet. The retweet feature in twitter helps users to quickly share the
Tweets with their followers, which can be from them or tweets from other users as well. Users
mention RT before the tweet to state that it is a retweet. In the present day, retweets signify the
importance or popularity of a particular tweet. The software collected the retweet counts within 1-
2 days of the posting of the tweets. The most retweeted tweet was posted by the username
“@GrrrGraphics” and received 1458 retweets. The tweet was, “What happens when left-wing
authoritarians run the most powerful companies”. The tweet suggests that it is related to Politics.
The second most retweeted tweet was by the same username “@GrrrGraphics” and received half
of the retweets as compared to the first one and got 713 retweets. The tweet read, “#Google Runs
over #FreedomOfSpeech #Censorship of #Conservatives #BenGarrison #cartoon
#JamesDamore”. Username “@clivebushjd” got 507 retweets with ‘#Censorship’. The tweet read,
“#Censorship Twitter Suspends Ohio #Republican Congressional Candidate- Chris Depizzo- After
Tweeting About #Democrat Rival”, which again is something related to Politics in United States
of America.
Tweets can be favorited by rolling over the tweet that needs to be favorited and the users can notice
a star show up on the right side of the tweet. The most favorited tweet contains 1993 favorites and
was regarded to the twitter support team. It read, “Dear @TwitterSupport You are successfully
censoring my account & continue blocking people from seeing my tweets!...”. The next in the line
are the ones which stayed on top of the most Retweeted list. The two tweets received 1931 and
1166 favorites respectively. The list of 13 most favorited tweets are given in the figure 5.
Figure 5: The most frequent used terms in the tweets
The most frequent used terms in the tweets were #Twitter, #Facebook, #Google, #YouTube etc.
The other most widely used terms included world, freedom, free speech, news, privacy, internet,
companies, conservative, authoritarian, people, movement, cartoon, conservative etc. Social
Media has been one of the widely used platforms when censorship is discussed which makes
Twitter and Facebook to stay in the list.
A total of 10,141 tweets were analyzed for sentiment analysis while excluding the retweets. In the
figure 6, which is meant to depict the sentiment analyses of the corpus of tweets that were collected
from Twitter. As shown, the number of negative tweets is enormously higher than the positive and
neutral sentiments with a total of 8,719 (85.98%) tweets. On the second place are the Neutral
sentiments of the users attached with the tweets, with 800 (7.89%) tweets and at the end is the
positive sentiments with 622 (6.13%) tweets related to the Tags “Censorship”. The tweets with
Negative sentiment have shown a series of ups and downs with fluctuation is tweet counts per day.
There could have been some serious issues with tags “#censorship” trending within the twitter
community on the 11th day when so many tweets were recorded with negative sentiments. Apart
from this, the Positive and Neutral tweets have stayed on the bottom competing with each other.
Figure 6: Sentiment Analysis
Conclusion
Millions of Tweets are created everyday by users around the world on the social networking site
Twitter to contribute their ideas, opinion, facts and news online. To understand this huge amount
of data present in texts through reading all or a part of it is very time taking and hectic. To utilize
the human’s visual interpretation and recognize easier ways for exploring the textual data,
information visualization tools are developed. These tools allow the users to understand data
visually on Twitter. While most of the existing tools are directed towards targeting few topics or
even a specific group of people, there has been a great demand for tools that are pointed at any
user who is interested in learning about the activities around the world through visualization over
Twitter.
Sentiment Analysis is growing at enormous rate due to the significance of automation in the fields
of mining, extracting and processing information in order to discover the general opinion of a
person. Sentiment Analysis is one way by which the researchers can explore Social Media
conversations along with other systematic viewpoint available for Knowledge Discovery. For
these reasons only, there is a need for further research to guide the researchers in order to select
and match various text analysis approaches using other social media data sources to generate
pinpoint and error-free outcomes. The use of sentiment analysis among various data analysis
method techniques is growing at a good pace. In this paper, we tried to investigate various
approaches towards sentiment analyses and data and text visualization.
There are some limitations to this research study. Firstly, the time period that was selected for the
research consisted of only one month, which led to the collection of a limited number of tweets by
the researchers. Secondly, the Tags used only one term, “#Censorship” and no related terms.
Censorship is a very broad topic which can be sub-divided into Book Censorship, Movie
Censorship, and News Censorship etc. There are related terms also such as Ban, Censor, and
Restrict etc. This study can be further enhanced by including related terms, increasing time span
etc.
References
• Ahlqvist, T., Back, A., Halonen, M., & Heinonen, S. (2008). Social Media Roadmaps:
Exploring the futures triggered by social media. Finland: Julkaisija. Available at:
http://www.vtt.fi/inf/pdf/tiedotteet/2008/t2454.pdf
• Bakliwal, A, et al. (2012). Mining sentiments from tweets. 3rd Workshop on Sentiment and
Subjectivity Analysis (WASSA) in Conjunction with 50th annual meeting of Association for
Computational Linguistics (ACL) 2012. Jeju, Island, Republic of Korea. Available at:
http://web2py.iiit.ac.in/research_centres/publications/download/inproceedings.pdf.a2ef40d7e7
aceb14.33325f50617065722e706466.pdf
• Balusamy, B., Varma, V. T. S., & Grandhi, S. S. M. Y. (2017). Social Network Web Mining:
Web Mining Techniques for Online Social Network Analysis. In Web Data Mining and the
Development of Knowledge-Based Decision Support Systems (pp. 284-310). IGI Global.
• Bello-Orgaz, G., Jung, J. J., & Camacho, D. (2016). Social big data: Recent achievements and
new challenges. Information Fusion, 28, 45-59.
• Kharde, V., & Sonawane, P. (2016). Sentiment analysis of twitter data: a survey of
techniques. International Journal of Computer Applications, 139(11), 5-15. Available at:
http://www.ijca.org/research/volume139/number 11/kharde-2016-ijca-908625.pdf
• Mostafa, M.M. (2013). More than words: Social networks’ text mining for consumer brand
sentiments. Expert Systems with Applications, 40(10), 4241-4251. Available at:
https://doi.org/10.1016/j.eswa.2013.01.019.
• Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations and Trends
in Information Retrieval, 2(1-2), 1–135. Available at:
http://www.cs.cornell.edu/home/llee/omsa/omsa.pdf
• Park, S., Ko, M., Kim, J. Liu, Y. & Song, J. (2011). The politics of comments: predicting
political orientation of news stories with commenters' sentiment patterns. In Proceedings of
the ACM 2011 conference on Computer supported cooperative work (CSCW '11). ACM:
New York, 113-122. Available at: https://doi.org/10.1145/1958824.1958842
• Ristoski, P., & Paulheim, H. (2016). Semantic Web in data mining and knowledge discovery:
A comprehensive survey. Journal of Web Semantics, 36, 1-22.
• Wani, M. A., & Jabin, S. (2018). Big Data: Issues, Challenges, and Techniques in Business
Intelligence. In Big Data Analytics (pp. 613-628). Springer, Singapore.
• Zhang L., Liu B. (2017) Sentiment Analysis and Opinion Mining. In: Sammut C., Webb
G.I. (eds) Encyclopedia of Machine Learning and Data Mining. Springer, Boston, MA.
Available at: https://doi.org/10.1007/978-1-4899-7687-1_907
