Sentiment Analysis on Educational Posts from

Social Media
Floradel S. Relucio Thelma D. Palaoag
Computer and Engineering Studies College of Information Technology
Department and Computer Science
Bicol University Polangui Campus University of Cordilleras
Polangui, Albay, Philippines Baguio City, Philippines
+639209269517 +639493666795
[email protected] [email protected]

ABSTRACT ideas are being expressed. From these, a research in sentiment

Social Networking on social media websites involves the use of the analysis can be made by automated extraction of positive and
internet to connect users with their friends, family and negative attitudes from text. Natural language processing and
acquaintances. Due to the increasing influence of social media such machine learning tools are being used to work with large volume
as Twitter, more number of users participates in the discussion and of text making it possible to extract from sentiments from social
different users belong to different kind of groups. Positive, media [1].
negative, and neutral comments are posted by the user and they In recent years, social media has grown extremely fast and in the
participate in the discussion. The study mainly focused on the vast number of posts being made and people express their opinions
educational posts gathered from the Twitter social media. The web about all kinds of topics. These opinions are very valuable and there
analytics technique was used in gathering and analyzing insights is a need for a way to automatically identify and extract them [2].
into community actions and attitudes through data collection, pre- As stated in the study of Yelena Mejova, “Social media offers a
processing, classification, and analysis of results. The collected powerful for people thoughts and feelings – it is an enormous ever-
tweets from February 1, 2017 to March 30, 2017 were 1,717 using growing source of texts ranging from everyday observations to
the keywords Philippine education, DepEd K-12 and CHED K-12. involved discussions [3].
After cleaning, 1,548 tweets were derived and classified as positive, The future of communication is social media. Social Media
negative, and neutral with 74.9% accuracy evaluation level. Results platforms and tools that enhances the experience of sharing
showed that mostly had expressed their negativity on the information. The platforms are the community-based input, content
implementation of the K-12 program in the country. Measures such sharing, contact and alliance. This interactive medium has a
as hiring of teachers, sufficient allocation and utilization of funds multitude of interesting things like forums, microblogging, social
for the procurement of books and resources, and sending the networking, social bookmarking, wikis and podcasts. It is the next
teachers concerned for training is deemed necessary to address the logical marketing arena and monitoring the social media activities
sentiments of the people. It is recommended then that an in-depth is a good way to measure customers’ loyalty and keeping a track on
study of the K-12 implementation may be conducted to further their sentiment towards brands or products. Currently, Facebook
improve the Philippine educational system. Results may also be dominates the digital marketing space, followed closely by twitter
presented to the officials of DepEd for validation and consideration [4].
to further enhance the implementation of K-12 program, as the next Social network tools provides opportunities to improve learning
step of the study. methods for both students and learning institutions. With the
networks, you can include media plugins that enable sharing and
CSS Concepts interaction. Students can benefit from social networks through
online tutorials and resources that are shared online. As cited in the
Information systems ➝ Data mining ➝ Data cleaning study of Abdelrazeq et al., A huge amount of sentiments are
expressed daily on University related topics using social media
Keywords platforms. Particularly, posted statements from students and
Sentiment analysis; social media; Twitter; K-12 program; teachers can provide a potential source for evaluating Universities
educational post [5]. Social media such as facebook, twitter among others is a
valuable source of knowledge. It is where one can gain new insights
1. INTRODUCTION on various topics or issues for academic purposes. It is also a
Many internet users actively use online review site, social networks medium where students can establish beneficial connections for
and personal blogs in expressing their opinions. Both positive and their careers, thus, it is imperative that an academic institution must
negative attitudes about people, places, organizations, events and be active in many social platforms that may provide training
strategies for students that will better shape their culture.
SAMPLE: Permission to make digital or hard copies of all or part of this
work for personal or classroom use is granted without fee provided that
The social network analysis in the frame of the negative links
copies are not made or distributed for profit or commercial advantage and prediction turns out to be useful in a vast number of areas, for
that copies bear this notice and the full citation on the first page. To copy example, recommendation systems in the social network. Since,
otherwise, or republish, to post on servers or to redistribute to lists, sites like Facebook, Twitter etc. do not provide with a feature of
requires prior specific permission and/or a fee. disliking a post of other user, but rather they let us write the
Conference’10, Month 1–2, 2010, City, State, Country. comments in relation to that post it seemed as a potent feature that
Copyright 2010 ACM 1-58113-000-0/00/0010 …$15.00. can be used for predicting the relations between the users [6].
DOI: http://dx.doi.org/10.1145/12345.67890
In the study of K.C. Khatib et al., it is stated that Social media is articles). Machine Learning was considered as one of the
playing a vital role in social networking and sharing of data. Social techniques or statistical analysis, but there has been little use of the
media is favored by many users as it is available to millions of fuzzy classifiers [12]. But as cited in the study of Bouazizi, et al,
people without any limitations to share their opinions and Sentiment analysis and opinion mining in social media is a hot topic
educational learning experience [7]. Social media are technologies of research. However, most of the collected researches from social
that facilitate social interaction, make possible collaboration and media and microblogging sites are oriented towards binary
enable deliberation across stakeholders. classification (i.e, classification into “positive” and “negative) or
In social media, every action you should be a part of a larger the ternary classification (i.e, classification into “positive”,
marketing strategy. That means every part, post, reply, like and “negative” and “neutral”) of texts [13].
comment in social networks should all be guided by a plan that’s Moreover, in the study of Rahamathulla et al., it stated they
exerting towards educational goals. S comprehensive social media developed a procedure for feature based sentiment analysis by
strategy might sound complicated but if it takes to create and the using a classifier called Support Vector Machine. This can be done
approach is correct, the rest of the social efforts follow naturally. by creating an automated system in which we can analyze the
As educators look for ways to engage and motivate students, social reviews posted by the users and extract the user’s perception about
media technologies are becoming a viable supplement to the a particular feature [14]. With the above-cited literatures, the
traditional learning environment. This research is needed to present study aimed to determine the sentiments of the people on
understand how the educational use of social media, impacts the current educational set-up of the country especially on the
student engagement and learning and to identify best practices and implementation of K-12 program.
guidelines for the use of social media in academic settings. The A conceptual model of the study was formulated as shown below.
study will collect different viewpoints on how social networking
sites can be integrated in education as well as in instruction and it
will be helpful to instructors, mentors and scholarly specialist in all
parts of training searching for a hypothetical functional way to deal
with creative educating. It was likewise featured the formal and
casual uses of social connection apparatuses as learning devices.
To make education convenient, the use of social media provides the
students with the ability to get more useful information to connect
with learning groups and educational institutions that adapt social
media developments into their systems and relying on group
resources and mechanisms that helps to improve the student life.
The rapid growth of online social media as twitter acts as a medium
where people contribute their opinion and emotions as text
messages. Opinion mining refers to the application of natural
language processing, computational linguistics, and text mining to
Figure 1. Conceptual Model of the Study
identify or classify whether the opinion expressed in text message
is positive or negative [8]. The messages include reviews and
opinions and so on. Hence, classifying the twitter educational posts
Figure 1 depicted that Educational posts from social media were
as to positive, negative, and neutral is considered in this study.
gathered first via Twitter. In gathering the said tweets, specific
Sentiment Analysis is a field that works on making sense out of
keywords will be used to easily collect data and will conduct pre-
textual feedback and opinions. As stated in the study of Aye et al.,
process and cleaning. Cleaning of data will be done in order to get
it says that Sentiment analysis is the analysis of the feeling such as
the authentic tweets and redundant tweets will be deleted. After
attitude, emotion and opinion [9]9. On social media, the sentiment
cleaning the data, classification of data will be annotated by
of a post can be seen in the tone or emotion conveyed in a brand
identifying the negative, positive and neutral tweets. Final process
mention. This has led to numerous data services, tools and analytics
represents the result of the analysis of the data and be able to
platforms. Some social media management tools often offers
determine educational posts that can be used for the improvement
algorithms or filters to make it easy to gauge sentiment. In social
of the Philippine educational system.
listening tools, sentiment analysis features will measure and report
on the tone or sentiment of your social mentions. Using social
networking outreach tools such as facebook, instagram, youtube
and twitter. With many types of social media, understanding 2. METHODOLOGY
customers’ opinion and subjectivity is regarded as an important
task in various domains and such opinions are propagated to other
users and might make a significant influence on them [10]. The web analytics technique was used to gather and analyze the
To fully understand the sentiment of short text the sentiment usage data to gain insights into community actions and attitudes.
structure is obtained from the dependency parsing with the Tools were made it possible to adopt a real data driven approach to
relationship and modified distance. It shows how to mitigate the understanding usage patterns to optimize the experience for the user
problem through sentiment structure and the sentiment calculation [15]. Anchored on the research methodology as shown in Figure 2,
rules [11]. the study involved several steps in conceptualizing and obtaining
In the study of Chris Jefferson, et al., it is stated that Sentiment the desired results.
analysis aims to identify the polarity of a document through natural
language processing, text analysis and computational linguistics.
Over the last decade, there has been much focus on sentiment
analysis as the data available on-line has grown exponentially to
include many sentiment based documents (reviews, feedback,
 each class. A trained data of 100 was set for each class as basis of
classifying the test data automatically. While the test data were
classified using automated classification tool, the 4j for Java, in
which each tweet collected was labeled by identifying most or 912
tweets as negative, 483 as positive and 153 as neutral.

Table 1. Sentiment Analysis Result

Annotation Sample Data Sets

Negative Back to Basics for Philippine Education !
Negative Department of Education : Rewrite Philippine
History books to include the full story of the
Martial Law era.
Negative Deped Be Like Kulang sa materials
Figure 2. Research Methodology
Negative Hey! Alam mo ba lahat na lang pinipilit ng
deped kahit di ready kaya maraming
napeperwisyo eh Ehem K-12 ehem Ncae ehem
2.1 Data Gathering Nat ano pa mga besh
Negative Pa k-12 pa pero pati materials di maayos ayos.
The educational posts were gathered from twitter social media yoqo na ncae deped yoqo na life
covering the periods from February 1, 2017 to March 30, 2017. Negative @DepEd _PH Ano po bang pwedeng
Using the GetOldTweets, from the time given, a total of 1,548 alternative program kung sakaling may failure
tweets were collected using the keywords “Philippine education”, ang K12 curriculum??? #IFailAtFBS
“DepEd K-12”, and “CHED K-12”.
Negative K12 READY PO BA TALAGA? SOBRANG
HIRAP MAPAG EKSPERIMENTUHAN!
2.2 Pre-Processing @DepEd _PH
Negative Bka pwedeng tanggalin na po ang k12 phirap
kc sobra!ako gsto ko mgaral ulit 1st yr college
As shown in Figure 3 is the acquired data that are not cleaned. pero jusme ang layo ng school @DepEd _PH
Before the data was prepared for sentiment analysis, manual
cleaning for the test data was made by deletion of redundant tweets. Negative K12 program is good program however we are
The three (3) labels of the trained data were however manually not ready for it hopefully #deped would see that
annotated. A total of 1,717 tweets were collected. After cleaning, Negative Pagsuspinde sa implementasyon ng K12 ng
there were 1,548 tweets that are used as data for sentiment analysis DepEd
classification.
Negative BKIT PINATUPAD ANG K12 KUNG
KULANG MGA CLASSROOMS? KULANG
CLASSROOM HANGGANG NGAYON NA
DEPED PINAKAMALAKING BUDGET
KADA TAON?
Negative DONT US! @DepEd _PH SNSBI NYO
HNDA KAYO? PRO NITEXTBOOKS OR
TEACHERS MNUAL AY WLA DHIL
INIUNA NYO YNG K12 ! THINK OF IT
@RRD_Davao @ABSCBNNews
Negative @DepEd _PH is true that some of the schools
sched for k12 are up to 7 pm?
Negative deped isnt ready. k12 gives great opportunities
but we're not ready for it. they implemented it
way too early
Figure 3. Sample Uncleaned Data
Negative Minamadali tayo ng DepEd . Hays. Di po kami
robot. Tao din po kami. Okay na sana yung K12
last year eh. Dinagdagan nyo pa. Hinay hinay
2.3 Classification lang.
Negative KAMI LANG BA MAY BAGONG
As shown in Table 1 is a sample of randomly picked common
GRADING SYSTEM NA YAN? HALOS
Educational tweets from the collected tweets that were classified as
LAHAT KAME BAGSAK SA FIRST
positive, negative, and neutral. A portion of the collected tweets
GRADING! @DepEd _PH #K12
were trained through manual annotations to build a trained set for
 Positive DepEd _PH will hire over 40 50 batch size, the accuracy of the three data tested based on the
three (3) classes obtained 74.9%.
Positive tama ka po Sir. Hindi po ma-reach out ng
Mostly expressed their negativity on our Philippine educational
DepEd yung iba. Pero sana magkaroon ng K12
system which implies that measures on the implementation of the
ang lahat ng mga public schools.
K-12 program needs further attention from its implementors such
Neutral To know more about K12 and its impact to as the Department of Education (DepEd), school heads, and senior
Higher Education high school teachers. Measures such as hiring of teachers,
Neutral Papayag po kaming isang buwan lang ang sufficient allocation and utilization of funds for the procurement of
vacation basta po tatanggalin niyo ang K12 books and resources, and sending the teachers concerned for
Deal po? training is deemed necessary to address the sentiments of the
people. It is recommended then that an in-depth study of the K-12
implementation may be conducted. Results may also be presented
to the officials of DepEd for validation and consideration to further
2.4 Results enhance the implementation of K-12 program, as the next step of
the study.
The final step is the result which represents the output of the
analysis of the data. Out of the 1,548 tweets, most or 912 tweets
were considered as negative, 483 as positive and 153 as neutral. 4. CONCLUSIONS
Most of the tweets which were classified as negative could however
be considered as significant considerations of the study for the
The use of social media in education provides students with
improvement of the Philippine educational system such as
learning groups and other educational system that make education
suspension of k-12, shortage of teachers, non-availability of books
convenient. Hence, classifying the twitter educational posts as to
and resources, and lack of teachers training. A positive feedback
positive, negative and neutral is considered in this study. Findings
was also noted that the Department of Education has budget
showed that mostly had expressed their negativity on the
allocation in hiring teachers and teachers training.
implementation of the K-12 program in the country.

It is thus that recommended that measures on the implementors of

3. RESULTS AND DISCUSSION the K-12 program such as DepEd, school heads, and senior high
school teachers to further improve the Philippine educational
system. Results may also be presented to the officials of DepEd for
Results showed that a total of 1,717 tweets were collected from validation and consideration to further enhance the implementation
February 1, 2017 to March 30, 2017 which dealt with educational of K-12 program, as the next step of the study.
posts using the keywords Philippine education, DepEd K-12, and
CHED K-12. After cleaning the data, URLs were removed since it
complicates the sentiment analysis process, Retweets or redundant 5. ACKNOWLEDGMENTS
tweets were also removed, repeated lettered are filtered. This
resulted to 1,548 data sets which were subjected for sentiment
analysis. The three (3) annotations were used in classifying: We would like to express our heartfelt thanks and appreciation to
positive, negative, and neutral. A trained data of 100 words was set all the people who extends support, shared unselfish insights,
for each annotation. suggestions and recommendations that enable our study complete
and most of all to God Almighty whose endless guidance and
blessings strengthened us and perseverance to make this study a
success.
Sentiment Analysis Result
6. REFERENCES
10% [1] Jayasanka, R.A.S.C., Madbushani, M.D.T., Marcus, E.R.,
Negative
Aberathne I.A.A.U. and Premaratne, S.C. 2014, “Sentiment
31% 59% Positive Analysis for Social Media. https://www.researchgate.net
[2] Lysedal, Tomas. 2014, “Sentiment Analysis of Swedish social
Neutral
Media”. https://www.researchgate.net
[3] Mejova, Yelena. 2012, “Sentiment Analysis within and across
social media systems. Ir.uiowa.edu
[4] Neri, Federico, Aliprandi, Carlo, Cuadros, Montserrat, Tomas.
2012, “Sentiemnt Analysis on Social Media”.
Figure 4. Percentage of Sentiment Analysis Result ieeexplore.ieee.org
[5] Abdelrazeq, Anas. 2015, Sentiment Analysis of Social Media
Figure 4 showed that out of the 1,548 tweets, the collected data for Evaluating Unversities”. https://scholar.google.com.ph
identified 912 tweets as negative which is fifty nine-percent (59%), [6] Das, Debasis and Sharma, Pushkar. 2017, “Algorithm for
483 as positive which is thirty one percent (31%) and 153 as neutral Prediction of Negative Links using Sentiment Analysis on
which is ten percent (10%) respectively. By setting 100 epochs and Social Networks”. https://www.researchgate.net
[7] Khatib, K.C., Kamble, T.D., Chendake, B.R., and Sonavane,
G.N. 2016, “Social Media Data Mining for Sentiment
Analysis”. www.irjet.net
[8] Vinodhini, G., and Chandrasekaran, RM. 2014, “Sentiment
Classification Using Principal Component Analysis Based
Neural Network Model”. ieeexplore.ieee.org
[9] Aye, Yu Mon, and Aung, Sinit. 2017. “Sentiment Analysis for
Reviews in Myanmar Text”. ieeexplore.ieee.org
[10] Trung, Nguyen Duc, Jung, Jason J., Yu, Le Anh and Kiss,
Attila. 2013, “Towards Modeling Fuzzy Propagation for
Sentiemnt Analysis in Online Social Networks: A Case Study
on Tweetscope”. ieeexplore.ieee.org
[11] Li, Jie and Oiu, Lirong 2017, Á Sentiment analysis method of
short texts in Microblog”. ieeexplore.ieee.org
[12] Jefferson, Chris, Liu, Han, and Cocea, Mihaela 2017, “Fuzzy
Approach for Sentiment Analysis”. ieeexplore.ieee.org
[13] Bouazizi, Mondher and Ohtsuki, Tomaski. 2017, Á Pattern-
Based Approach for Multi-Class Sentiment Analysis in
Twitter”. ieeexplore.ieee.org
[14] Rahamathulla, Hari Krishna M., and Akbar Ali. 2017,
“AFeature Based Approach for Sentiment Analysis using SVM
and Conference Resolution”. ieeexplore.ieee.org
[15] Ali, Anwaar, 2016, “Big Data for Development Applications
and Techniques”. https://scholar.google.com.ph