ArticlePDF Available

Abstract and Figures

COVID pandemic has impacted most of the higher education institutes by shifting to online teaching. Due to this shift, there has been increasing use of e-text, online learning management systems, social media apps, and micro-blogging platforms used by students to provide feedback and comments about a course and online classroom experience before and during the pandemic. This feedback is important for the education institutes and can be used to improve the teaching and learning experience. One of the major problems is to extract useful information out of several comments and feedback. This study presents a hybrid approach of student sentiment analysis based on feedback of classes collected through Google survey forms and WhatsApp social media platforms before and during the pandemic. For classification and comparative analysis, Support Vector Machine, and Naïve Bayes algorithms have been used and an average accuracy of 85.62% has been achieved using Support vector machine with K-fold cross-validation.
Content may be subject to copyright.
Umair et al., International Journal on Emerging Technologies 12(2): 177-182(2021) 177
International Journal on Emerging Technologies 12(2): 177-182(2021)
ISSN No. (Print) : 0975-8364
ISSN No. (Online) : 2249-3255
Sentiment Analysis of Students’ Feedback before and after COVID-19 Pandemic
M. Umair1, A. Hakim1*, A. Hussain1 and S. Naseem2
1Researcher, Department of Computer Science, MNS University of Agriculture, Multan, Pakistan.
1Assistant Professor, Department of Computer Science, MNS University of Agriculture, Multan, Pakistan.
1Assistant Professor, Department of Computer Science, MNS University of Agriculture, Multan, Pakistan.
2Lecturer, Department of Statistics, MNS University of Agriculture, Multan, Pakistan.
(Corresponding author: A. Hakim)
(Received 05 April 2021, Revised 26 May 2021, Accepted 22 June 2021)
(Published by Research Trend, Website: www.researchtrend.net)
ABSTRACT: COVID pandemic has impacted most of the higher education institutes by shifting to online
teaching. Due to this shift, there has been increasing use of e-text, online learning management systems,
social media apps, and micro-blogging platforms used by students to provide feedback and comments about
a course and online classroom experience before and during the pandemic. This feedback is important for
the education institutes and can be used to improve the teaching and learning experience. One of the major
problems is to extract useful information out of several comments and feedback. This study presents a
hybrid approach of student sentiment analysis based on feedback of classes collected through Google
survey forms and WhatsApp social media platforms before and during the pandemic. For classification and
comparative analysis, Support Vector Machine, and Naïve Bayes algorithms have been used and an average
accuracy of 85.62% has been achieved using Support vector machine with K-fold cross-validation.
Keywords: COVID-19, Opinion mining, Sentiment analysis, online feedback, Natural language processing, COVID.
Abbreviations: COVID, Coronavirus Disease; AI, Artificial Intelligence; ML, Machine Learning; LMS, Learning
Management System; LIWC, Linguistic Inquiry Word Count; NB, Naïve Bayes; MNB, Multinomial Naïve Bayes; RF,
Random Forest; SGD, Stochastic Gradient Descent; SVM, Support Vector Machine; NLTK, Natural Language Toolkit;
VADER, Valence Aware Dictionary and Sentiment Reasoner; NLP, natural language processing; GSP, Generalized
Sequential Pattern; TD-IDF, Term Frequency-Inverse Document Frequency; BOW, Bag of Words.
I. INTRODUCTION
Sentiment Analysis, also known as Opinion mining or
Opinion AI is one of the leading research areas due to
the increasing use of e-text, online learning
management system (LMS), and microblogging
platforms including Twitter, WhatsApp, Facebook,
AnswerGarden, Poll Everywhere, and online blogs.
People use these platforms to express their thoughts
and opinion regarding any product, place, or event
throughout the world. This feedback is very critical for
concerned organizations to satisfy their customers. One
of the major problems is to extract useful information out
of feedback and comments to improve the quality of
products or services.
Sentiment Analysis is the process of classification of
data in different categories such as positive, negative, or
neutral class. In this work, we gathered student
feedback in two different scenarios of before and during
the COVID pandemic from WhatsApp groups and
Google forms. An analysis has been performed on
student’s textual comments in both scenarios using
sentiment analysis to identify their opinion about the
quality of teaching and learning. One of the objectives of
this system is to help university faculty and
administration to find gaps between students learning
and tutor teaching quality.
Sentiment analysis is mostly done using three
techniques: lexicon-based approach, machine learning
approach, and a hybrid technique that uses both lexicon
and machine learning techniques [1, 2]. In the lexicon-
based technique, a pre-defined dictionary is used where
words are already weighted according to their
sentiments. The most common tools for performing
lexicon-based analysis are SenticNet, SentiStrength,
and Linguistic Inquiry Word Count (LIWC) [3]. In the
machine-learning approach, the common algorithms for
sentiment classification are Naïve Bayes (NB),
Multinomial Naïve Bayes (MNB), Random Forest (RF),
Stochastic Gradient Descent (SGD), Support vector
machine (SVM), and hybrid techniques using a
combination of these algorithms.
Due to the COVID-19 pandemic situation, almost all
higher education institutes were shifted from face-to-
face or blended to fully online teaching, and many
students faced difficulties regarding online learning. To
facilitate them, it is critical to design a system that takes
feedback from students and provide an overall class
opinion to teachers and administrators to revise their
teaching and assessment methodology. This can be
done by converting textual data into positive, negative,
and neutral classes using Natural language processing
algorithms. It will help to determine students’ level of
understanding and any difficulties they are facing after
each class and to classify text-based data into positive,
negative, and neutral feedback to evaluate tutor
performance.
In this work, we have performed classification and
analysis of student’s feedback before and after COVID-
19 pandemic using machine learning techniques. Data
is collected through online google forms, LMS, and
e
t
Umair et al., International Journal on Emerging Technologies 12(2): 177-182(2021) 178
Whatsapp group messages of selected courses.
Annotation of student’s feedback on classes before and
during pandemic has been performed using open-
source tools: TextBlob and VADER [4]. These tools
were integrated with the Natural Language Toolkit
(NLTK) Python library using Jupiter notebook.
VADER (Valence Aware Dictionary and sEntiment
Reasoner) is a rule-based and vocabulary-based
sentiment analysis tool adapted to the sentiment
expressed on social media and other types of data [5].
VADER is different from other emotional analysis tools
because it does not classify text in discrete categories of
positive, negative, or neutral. Instead, VADER
generates a composite score between -1 and +1, which
indicates a range of positive, negative, or neutral.
TextBlob is a Python library for processing textual data
that provides a simple Application programming
interface to perform common natural language
processing (NLP) tasks like speech part tagging, noun
phrase extraction, emotional analysis, classification, and
translation. It generates a result score between -1 and
+1. Score grading for both TextBlob and VADER is such
that all values less than zero are considered as
negative, zero consider neutral and values above zero
considered as positive. For classification, Naïve Bayes,
and Support Vector Machine [6] have been used due to
their high accuracy on textual data [7].
This paper is divided into five sections. Section 2
presents the background to the related literature of
sentiment analysis and opinion mining. It also discusses
the main approaches and techniques for sentiment
classification. The proposed methodology is described
in Section 3. Results and discussions are presented in
Section 4 followed by a conclusion and future work that
is discussed in Section 5.
Sentiment analysis of text data on the Internet can be
useful in many applications such as product reviews,
political campaigns, popular topics, and educational
improvements, but it is most often applied to feedback
analysis [8]. Feedbacks are received from several
domains, including advertising [9] , movies [3, 10, 11,
12], products [4, 10, 13], automobiles, tourism,
smartphones [7, 14] , and education-learning [15-17].
We focus on an overview of previous work related to the
sentimental analysis of educational related data.
An opinion is someone’s feelings, beliefs, or judgments
about an important issue in a particular situation and is
generally considered subjective. Studies have shown
that opinions have a major impact not only on facts but
also on individual decisions, as well as on communities,
such as organizations and government sectors. The
terms sentiment analysis and opinion mining often used
interchangeably in the field of text data mining which
extracts opinions from evaluation texts and classifies the
polarity of opinions into positive or negative rankings
based on the valency of the text results [18].
In the field of education, very few studies have focused
on sentiment analysis of online learning such as [19-23],
and only one feedback study was conducted on the
classroom related data [24]. In [25], authors used two
pattern mining algorithms Apriori and Generalized
Sequential Pattern (GSP) to extract comment words
from student feedback data sets for evaluation of tutor
performance. The results of the evaluation showed that
GSP performs better than Apriori in the extraction of
opinion words.
In [26], authors claimed to detect polarity (positive,
negative, and neutral) on Facebook comments related
to e-learning. The positive class consisted of happy and
excited emotions, and negative class included emotions
such as anger, sadness, and profanity. The method was
evaluated on 1000 positive and negative text messages
and Facebook statuses.
Most of the sentiment analysis techniques focused on
after-class feedback and classified students’ feedback
into discrete categories. In this work, we are performing
comparative analysis of student’s feedback in blended
and fully online learning in two scenarios of before and
after COVID pandemic respectively.
II. MATERIALS AND METHODS
The proposed framework for this research (as shown in
Fig. 1) consists of five phases: Dataset collection and
preprocessing, data annotation/polarity calculation,
feature extraction, classification, and visualization
Fig. 1. Proposed Framework for Sentiment Analysis.
A. Data Collection
For data collection purposes three different sources
were used. The first was before the COVID-19
pandemic, where classes were conducted in traditional
or blended mode. Feedback of undergraduate and
postgraduate students was collected during and after
each class of the selected courses from Department of
Computer Science at MNS-University of Agriculture. A
google questionnaire form is designed with the
coordination of the Supervisor. The others source was
feedback collected using an online Google form
provided to students at the end of the class.
During COVID-19 pandemic scenario, all classes were
shifted to fully online mode. We extracted feedback from
students’ WhatsApp groups (see Appendix-I for details).
The students sent textual comments/messages
expressing their opinions and feedback about the
lecture. An un-labelled dataset of around 2000
instances was collected from all sources. Word Cloud of
all words in the dataset is shown in Fig. 2.
Umair et al., International Journal on Emerging Technologies 12(2): 177-182(2021) 179
Fig. 2. Word cloud of the dataset.
There were several challenges in classification of
sentiments for instance, the data collected from
Whatsapp messages was unstructured that made it
hard to annotate and classify data using machine
learning techniques. Students’ feedback data was noisy,
for instance, the feedback “Today's lessons are
understandable and the topic is awesome”. There were
several deviations of spelling of the same word,
awesome word can be seen in many different ways as-
“awsem, awssuumm, awsomee”. The process of data
cleaning is discussed in the next section.
Pre-processing
Input data pre-processing is a significant step before the
classification process. At this stage, the dataset is first
normalized and then prepared for the classification
algorithm training, so that proposed algorithms work
efficiently and achieve actual results consuming less
time [32]. We have used pre-processing parameters:
stemmer, stop-words handler, lowercase or uppercase
conversion, and tokenizer [9, 27, 28, 29], for cleaning up
junk data and increasing data accuracy by reducing
data errors. There are pros and cons of pre-processing,
such as without pre-processing, the system may lose
the importance of words, while on the other hand, the
loss of important data might occur due to the extensive
pre-processing. An open-source web application Jupyter
Notebook has been used for data statistical modelling,
numerical simulation, data visualization, data cleaning
and transformation, and machine learning model
implementation. Jupyter Notebook Anaconda 3 has
been used for data pre-processing and model
implementation using python language.
Data Annotation
Polarity detection of textual data has been performed
using two methods: TextBlob and VADER. TextBlob
annotated our total dataset into 650 negative feedbacks,
900 neutral feedbacks, and 450 positive feedbacks as
shown in Table 1. The problem with TextBlob
annotations is that if feedback received from a student is
in the form of ‘yes’or ‘no’ for any question, instead of
annotating it as ‘positive’ or ‘negative’ based on the
context, TextBlob annotates it as ‘neutral’. To perform
context-aware annotation, we used Vader sentiment
analysis that annotated our total dataset into 633
Negative feedbacks, 793 Neutral feedbacks, and 574
positive feedbacks as shown in Table 2.
Table 1: Dataset annotation using TextBlob.
Before COVID
During COVID
Class
Class
Negative 160 Negative 490
Neutral 400 Neutral 500
Positive 240 Positive 210
Total 800 Total 1200
Table 2: Dataset annotation using Vader.
Before COVID
During COVID
Class
Class
Negative 134 Negative 499
Neutral 450 Neutral 343
Positive 323 Positive 251
Total 800 Total 1200
Feature Extraction
annotating the dataset using two different techniques,
we applied vectorization, Term Frequency – Inverse
Document Frequency TF-IDF, and Bag of word
techniques for feature extraction.
Vectorization: As we know Machine Learning (ML)
algorithms operate best on numeric values, where rows
represent instances and columns represent features into
two-dimensional feature matrices. To perform ML in
text, the document needs to be converted into a vector
representation to apply numerical machine learning. The
numerical representation of the document gives the
ability for performing meaningful analysis of ML
algorithms.
Term Frequency-Inverse Document Frequency (TF-
IDF): In the pre-processing phase, TF-IDF provides a
numerical statistic to state the importance of a word in a
document that helps in sentiment analysis (Pang & Lee,
2008). Frequency evaluation of commonly use words
has been performed in the TF-IDF phase to identify
important words in the dataset. In the dataset, the most
frequently used words are ‘yes’, ‘no’, ‘excellent’, ‘bad’,
‘sad’, or ‘happy’. The number of occurrences of a
term/word in a given document is referred to as term
frequency.
Umair et al., International Journal on Emerging Technologies 12(2): 171-176(2021) 180
Bag of words (BOW): Extracting information is a key
phase in text mining, which serves as the starting point
in many data mining algorithms. Extracting entities and
their relationships from the text may reveal meaningful
semantic information in the text data rather than a
general representation of the text bag, and it is
necessary to understand the hidden knowledge in the
text data. For feature extraction, we have used a
common technique called a bag of word approach
(BOW). Using this approach, a list of unique words in
the dataset is created which is referred to as
vocabulary. This approach performs analysis of the
histogram of words within the text by considering each
word as a characteristic assign a value of 1 if the word
is present and 0 if the word is absent in the vector
representation of data.
Classification
After feature extraction, we used Naïve Bayes, and a
Support vector machine (SVM) for classification. Both
are the famous machine learning algorithms for
supervised based approach and in the literature, both
are followed by different researchers [7, 30] for
classification purpose. Both perform batter accuracy for
text classification rather than other supervised machine
learning classifiers after comparisons [31].
SVM classifier works best for classifying sparse text
data by defining rectilinear partitions in the data set and
divides the set into different classes. SVM also uses
kernel functions to transform data in a certain way, so
that Hyperplane distribution classes can be allocated
efficiently. On the other hand, Naïve Bayes classifier is
the most commonly used text mining classifier which
uses Bayes theorem to calculate the possibility of the
given label related to a particular feature
III. RESULTS AND DISCUSSION
To analyze the classifier’s performance with a sentiment
analysis framework using both TextBlob, and Vader
annotated dataset, three measures have been taken
(Recall, Precision, and F-Measure) to evaluate the
performance of the proposed method.
TextBlob Annotated dataset classification results
To optimize results and overcome the model overfilling
problem we work on k-fold cross-validation in this model
we calculate the best value of K by using hyper-
parameter tuning using Grid Search for maximum
results for both TextBlob annotated dataset and Vader
annotated dataset, In this validation we use K as 5
which provides maximum result scores after applying
hyper-parameter tuning using Grid Search technique on
TextBlob annotated dataset and analyze the results
which are the mean scores for recall, precision, and F-
Measure score are 70.0%, 71.0%, and 69.0% with
70.8% accuracy using SVM and 69.8% accuracy by
using Naïve Bayes, see Table. 3 class wise (negative,
neutral, positive). Figure 3 shows classification results
after TextBlob annotation through pie chart showing
63% neutral responses, 29% positive (high positive +
positive), and 8% negative (high negative + negative)
responses in the entire dataset.
Fig. 3. Classification results of TextBlob annotated data.
Table 3: Results for TextBlob Annotated Dataset
using train-test split using SVM.
Class
Precision
Recall
F
-
Measure
Negative
0.61 0.72 0.70
Neutral
0.50 0.57 0.53
Positive
0.71 0.69 0.70
Vader Annotated dataset classification results
One of the issues of using TextBlob annotation is that if
feedback received from the student says only ‘Yes’ or
‘No’ for any question, instead of annotating it as positive
or negative it shows that the given feedback is ‘neutral’.
That’s why the neutral value graph also shows more
percentage. To deal with this problem we move forward
to Vader sentiment analysis whose annotation result is
more authentic as compared to TextBlob. To optimize
results and overcome the model overfitting problem we
work on k-fold cross-validation in this model we use the
value of k as 15 which provides maximum result scores
after applying hyper-parameter tuning using Grid Search
technique using both algorithms (SVM and Naïve
Bayes). The obtained results are mean scores of 73.3%
accuracy using Naïve Bayes, and 85.6% accuracy using
Support vector machine which is higher than Naïve
Bayes. According to the results the Precision, Recall,
and F-Measure of Support vector machine is 85.0%,
83.0%, and 82.0%. The results are presented class-
wise (negative, neutral, positive) as shown in Table 4.
Table 4: Results for TextBlob Annotated Dataset using train-test split using SVM.
Class Precision Recall F-Measure
Negative 0.71 0.99 0.83
Neutral 0.55 0.60 0.63
Positive 0.97 0.94 0.53
7%
22%
1%
7%
63%
High Positive Positive High Negative
Negative Neutral
Umair et al., International Journal on Emerging Technologies 12(2): 171-176(2021) 181
Fig. 4 shows pie chart of all Vader annotated feedback
after classification showing 42% neutral responses, 41%
positive, and 17% negative responses in the entire
dataset.
Fig. 4. Pie chart of all polarity values (annotated using
Vader).
Fig. 5. Pie chart of manual annotation.
After TextBlob and Vader, we also annotated our
dataset based on expert opinion of three human
experts. Fig. 5 shows that human experts assign 54%
positive responses, 19% neutral responses, and 27%
negative responses in the entire dataset.
A comparison table to compare annotation percentage
of TextBlob, VADER sentiment analysis, and manual
annotation is shown in Table 5. The difference between
TextBlob, VADER, and manual annotation polarity
distribution is because of TextBlob and VADER use
different libraries and lexicons for the polarity distribution
process. VADER sentiment analysis is most applicable
to social media and even educational text. It relies on a
dictionary of emotional words. Each word in the
dictionary is divided into positive or negative numbers.
Table 5: Comparison of dataset annotation using
different methods.
Annotation
Methods
Positive
comments
(%)
Negative
comments
(%)
Neutral
comme
nts (%)
TextBlob 29 8 63
VADER 41 17 42
Manual
Annotation 54 27 19
IV. CONCLUSION
The purpose of this research is to propose a
comparative analysis of student’s feedback before and
during the COVID pandemic when all educational
institutes have been shifted from face-to-face learning to
a fully online learning system. Student’s feedback has
been collected through different online platforms
including WhatsApp and google forms. We applied both
Naïve Bayes and Support vector machine supervised
machine learning algorithms for classification and
compared their performance on the given dataset.
Support vector machine works best for text-polarity
classification in our study. It was found that there are
more negative instances of feedback during fully online
classes as compared to that in the blended teaching
mode (Table 2). The findings can help tutors in
designing strategies for teaching improvement in fully
online classes after each class.
V. DISCUSSION &FUTURE SCOPE
Due to COVID-19 pandemic, most of the universities
around the world have shifted to blended learning or
fully online learning systems. In this scenario, there is a
need to get informed about the students’ sentiments and
their opinion when the student-teacher interaction is
minimal. There is a potential to merge the proposed
system with real time facial expression analysis to
develop an information board for the tutors. The results
presented in their paper shows the performance of
different annotation methods and classification
techniques for sentiment analysis out of textual data.
Based on results, it is recommended to use VADER for
annotation and SVM for classification of textual data.
We are also working on automated detection of
learner’s engagement by analyzing their facial
expressions. We plan to compare the results of both
systems to find the difference (if any) between the
written (through comments/text messages/forms) and
expressed (through face) sentiments of the students.
The comparative analysis could be applied in the quality
assurance program at the university to determine and
improve the student-tutor relationship.
Conflict of Interest: There is no conflict of interest
relevant to publication of this paper.
REFERENCES
[1]. Pang, B., & Lee, L. (2009). Opinion mining and
sentiment analysis. Comput. Linguist, 35(2), 311-312.
[2]. Saif, H., He, Y., Fernandez, M., & Alani, H. (2016).
Contextual semantics for sentiment analysis of Twitter.
Information Processing & Management, 52(1), 5-19.
[3]. Ahmad, M., Aftab, S., Muhammad, S. S., &
Waheed, U. (2017). Tools and techniques for lexicon
41%
17%
42%
Positive Negative Neutral
54%
27%
19%
Positive
Negative
Neutral
Umair et al., International Journal on Emerging Technologies 12(2): 177-182(2021) 182
driven sentiment analysis: a review. Int. J. Multidiscip.
Sci. Eng, 8(1), 17-23.
[4]. Elli, M. S., & Wang, Y.-F. (2016). Amazon Reviews,
business analytics with sentiment analysis.
Hutto, C., & Gilbert, E. (2014, May). Vader: A
parsimonious rule-based model for sentiment analysis
of social media text. In Proceedings of the International
AAAI Conference on Web and Social Media (vol 8(1)).
[6]. Cortes, C., &Vapnik, V. (1995). Support-vector
networks. Machine learning, 20(3), 273-297.
[7]. Ahmad, M., Aftab, S., Bashir, M. S., & Hameed, N.
(2018). Sentiment Analysis using SVM: A Systematic
Literature Review. International Journal of Advanced
Computer Science and Applications, 9(2), 182-188.
[8]. Jagtap, V., &Pawar, K. (2013). Analysis of different
approaches to sentence-level sentiment classification.
International Journal of Scientific Engineering and
Technology, 2(3), 164-170.
[9]. Altawaier, M.M., & Tiun, S. (2016). Comparison of
machine learning approaches on arabic twitter
sentiment analysis. International Journal on Advanced
Science, Engineering and Information Technology, 6(6),
1067-1073.
[10]. Ahmad, M., Aftab, S., & Ali, I. (2017). Sentiment
analysis of tweets using svm. Int. J. Comput. Appl.,
177(5), 25-29.
[11]. Asur, S., & Huberman, B. A. (2010). Predicting the
future with social media. Paper presented at the 2010
IEEE/WIC/ACM international conference on web
intelligence and intelligent agent technology.
[12]. Pang, B., Lee, L., & Vaithyanathan, S. (2002).
Thumbs up? Sentiment classification using machine
learning techniques. Paper presented at the
Proceedings of the ACL-02 conference on Empirical
methods in natural language processing-Volume 10.
[13]. Thelwall, M., & Buckley, K. (2013). Topicbased
sentiment analysis for the social web: The role of mood
and issuerelated words. Journal of the American
Society for Information Science and Technology, 64(8),
1608-1617.
[14]. Zavvar, M., Rezaei, M., & Garavand, S. (2016).
Email spam detection using combination of particle
swarm optimization and artificial neural network and
support vector machine. International Journal of Modern
Education and Computer Science, 8(7), 68.
[15]. Qiuhan, L., Afzaal, M., Alaudan, R., & Younas, M.
(2020). COVID 19 Pandemic and Online Education in
Hong Kong: An Exploratory Study. International Journal
on Emerging Technologies, 11(5), 411-418.
[16]. Arshad, M., Almufarreh, A., Noaman, K. M., &
Saeed, M. N. (2020). Academic Semester Activities by
Learning Management System during COVID-19
Pandemic: A Case of Jazan University. International
Journal on Emerging Technologies, 11(5), 213-219.
[17]. Poulos, A., &Mahony, M. J. (2008). Effectiveness
of feedback: The students’ perspective. Assessment &
Evaluation in Higher Education, 33(2), 143-154.
[18]. Medhat, W., Hassan, A., & Korashy, H. (2014).
Sentiment analysis algorithms and applications: A
survey. Ain Shams engineering journal, 5(4), 1093-
1113.
[19]. Altrabsheh, N., Cocea, M., & Fallahkhair, S.
(2014b). Sentiment analysis: towards a tool for
analysing real-time students’ feedback. Paper presented
at the 2014 IEEE 26th international conference on tools
with artificial intelligence.
[20]. Aung, K.Z., & Myo, N.N. (2017). Sentiment
analysis of students' comment using lexicon-based
approach. Paper presented at the 2017 IEEE/ACIS 16th
International Conference on Computer and Information
Science (ICIS).
[21]. Chamlertwat, W., Bhattarakosol, P., Rungkasiri, T.,
& Haruechaiyasak, C. (2012). Discovering Consumer
Insight from Twitter via Sentiment Analysis. J. UCS,
18(8), 973-992.
[22]. Jing, L., & Yanqing, Z. (2012). Teaching evaluation
method based on least squares support vector machine
and chaos particle swarm optimization algorithm.
JDCTA: International Journal of Digital Content
Technology and Its Applications, 6, 343-351.
[23]. Patel, T., Undavia, J., & Patela, A. (2015).
Sentiment analysis of parents’ feedback for educational
institutes. International Journal of Innovative and
Emerging Research in Engineering, 2(3), 75-78.
[24]. Altrabsheh, N., Cocea, M., &Fallahkhair, S.
(2014a). Learning sentiment from students’ feedback for
real-time interventions in classrooms. Paper presented
at the International Conference on Adaptive and
Intelligent Systems.
[25]. Rashid, A., Asif, S., Butt, N. A., & Ashraf, I. (2013).
Feature level opinion mining of educational student
feedback data using sequential pattern mining and
association rule mining. International Journal of
Computer Applications, 81(10).
[26]. Ortigosa, A., Martín, J. M., &Carro, R. M. (2014).
Sentiment analysis in Facebook and its application to e-
learning. Computers in human behavior, 31, 527-541.
[27]. Jeong, H., Shin, D., & Choi, J. (2011). Ferom:
Feature extraction and refinement for opinion mining.
Etri Journal, 33(5), 720-730.
[28]. Hu, M., & Liu, B. (2004). Mining and summarizing
customer reviews. Paper presented at the Proceedings
of the tenth ACM SIGKDD international conference on
Knowledge discovery and data mining.
[29]. Kobayashi, N., Inui, K., & Matsumoto, Y. (2007).
Opinion mining from web documents: Extraction and
structurization. Information and Media Technologies,
2(1), 326-337.
[30]. Yadav, S. K. (2015). Sentiment analysis and
classification: a survey. International Journal of Advance
Research in Computer Science and Management
Studies, 3(3), 113-121.
[31]. Shivaprasad, T. K., & Shetty, J. (2017, March).
Sentiment analysis of product reviews: a review. In 2017
International Conference on Inventive Communication
and Computational Technologies (ICICCT) (pp. 298-
301). IEEE.
[32]. Zainudin, S., Jasim, D. S., & Bakar, A. A. (2016).
Comparative analysis of data mining techniques for
Malaysian rainfall prediction. Int. J. Adv. Sci. Eng. Inf.
Technol, 6(6), 1148-1153.
How to
cite this article
:
Umair, M., Hakim, A., Hussain, A. and Naseem, S.
(2021).
Sentiment Analysis of Students’
Feedback before and after COVID-19 Pandemic. International Journal on Emerging Technologies, 12(2): 177–182.
... For example, you could identify frequently used positive words in positive sentiment comments and vice versa (Kastrati et al., 2021). Additionally, you could analyze the sentiment associated with specific keywords related to online learning aspects like "flexibility," "interaction," "difficulty," and more (Umair et al., 2021). The last step is to interpret the findings to understand the overall sentiment of students toward online learning and to identify specific aspects that contribute to positive or negative sentiment (Mendoza, Yan, & King, 2023). ...
... While challenges exist, the positive sentiment surrounding convenience and flexibility suggests that online learning is a viable and valued educational option. To enhance the online learning experience, institutions should address specific pain points identified in the analysis and continue to adapt their online offerings to meet the diverse needs of their students (Umair et al., 2021;Wong, 2023). ...
Article
The post-pandemic situation has triggered a seismic shift in the education landscape, forcing educational institutions worldwide to transition to online learning platforms. Understanding students' perspectives towards this abrupt transformation is crucial for improving the quality of online education. This study employs sentiment analysis techniques, specifically Word Frequency Analysis, to gain insights into students' sentiments and opinions concerning online learning. Using a dataset of student responses collected through surveys and open-ended questions, the researchers conducted a comprehensive analysis of the words and phrases frequently used by students when describing their experiences with online learning. Through the systematic identification of prominent words and their frequencies, this research provides a valuable overview of the prevailing sentiments, concerns, and areas of satisfaction among students. The findings reveal a nuanced picture of students' attitudes. While positive sentiments such as “nyaman”, “baik”, and “responsif" are prevalent, there are also concerns and challenges highlighted, including 'isolation,' 'technical issues,' and 'communication difficulties.' Furthermore, the analysis uncovers variations in sentiment based on demographic factors, course types, and prior experience with online education. This research contributes to a deeper understanding of students' sentiments towards online learning, offering educators and policymakers valuable insights for enhancing the online learning experience. The Word Frequency Analysis methodology employed in this study proves to be a powerful tool for efficiently distilling the sentiments expressed in large datasets of qualitative responses, providing a foundation for further qualitative and quantitative investigations in the field of online education.
... For the 5-Fold cross-validation technique, the SVM outperforms the NB with an accuracy of 90.17%. Umair & Hakim (2021) developed a sentiment analyser model for students' feedback before and after the COVID-19 pandemic. The SVM algorithm utilised performed better than the NB with an accuracy of 85.62%. ...
... The SVM machine learning algorithm has shown promising results compared to other classification algorithms for sentiment analysis in evaluating students' feedback. Even though some literature (Lalata et al., 2019) and (Umair & Hakim, 2021) utilised feature selection mechanisms for sentiment analysis, it resulted in low classification accuracy. ...
Article
Full-text available
Text analytics in education has evolved to form a critical component of the future SMART campus architecture. Sentiment analysis and qualitative feedback from students is now a crucial application domain of text analytics relevant to institutions. The implementation of sentiment analysis helps understand learners' appreciation of lessons, which they prefer to express in long texts with little or no restriction. Such expressions depict the learner's emotions and mood during class engagements. This research deployed four classifiers, including Naïve Bayes (NB), Support Vector Machine (SVM), J48 Decision Tree (DT), and Random Forest (RF), on a qualitative feedback text after a semester-based course session at the University of Education , Winneba. After enough training and testing using the k-fold cross-validation technique, the SVM classification algorithm performed with a superior accuracy of 63.79%.
... Through this technological tool, authorities can apply corrections or provide feedback measures to enhance the academic service offered. In the same line of thought, [29] and [30] point out that the use of text messaging through the WhatsApp mobile application is beneficial for communication between students and teachers. It serves not only as a means of information, but also as a tool for organizing and actively participating in academic or research projects, thereby promoting teamwork. ...
Article
Full-text available
In the context of virtual teaching during the COVID-19 pandemic, a gap emerged between students and teachers due to social distancing measures. This gap hindered the flow of information about the teaching-learning process, making it difficult for authorities to make informed decisions to improve student satisfaction and teaching performance. In this context, the widespread use of mobile applications, through which students express their opinions on the conditions of their learning sessions, is significant. In this sense, the objective of this paper is to apply voice analytics to identify the factors that contribute to the lowest level of student satisfaction in teacher performance using WhatsApp audio messaging. The study has a quantitative approach, an exploratory-descriptive level, and a non-experimental cross-sectional design. The study population consisted of 33 students. It was determined that the factor with the lowest level of satisfaction is the dimension “class session administration,” with a percentage of 57.58%, which is significantly lower than the satisfaction levels of the other factors analyzed, which are above 90%. Therefore, it is concluded that in addition to using rubrics to evaluate teacher performance in adhering to lesson plans and class sessions, the authorities should also implement regulations that support the use of voice analytics through mobile applications like WhatsApp. This will provide insights from students, who are direct participants in the teaching process, regarding their perception of teaching performance.
... Accuracy (ACC), precision (P), recall (R), and F-measure (F) values are employed as performance measures in this study. The experiment is modelled with the support of machine learning Naive Bayes, the most common text mining classifier which uses Bayes theorem to calculate the possibility of the given label related to a particular feature [70]. Moreover, Naive Bayes is used due to their high accuracy on textual data [71]. ...
Article
Full-text available
In Malaysia, during the early stages of the COVID-19 pandemic, the negative impact on mental health became noticeable. The public's psychological and behavioral responses have risen as the COVID-19 outbreak progresses. A high impression of severity, vulnerability, impact, and fear was the element that influenced higher anxiety. Social media data can be used to track Malaysian sentiments in the COVID-19 era. However, it is often found on the internet in text format with no labels, and manually decoding this data is usually complicated. Furthermore, traditional data-gathering approaches, such as filling out a survey form, may not completely capture the sentiments. This study uses a text mining technique called Latent Dirichlet Allocation (LDA) on social media to discover mental health topics during the COVID-19 pandemic. Then, a model is developed using a hybrid approach, combining both lexicon-based and Naïve Bayes classifier. The accuracy, precision, recall, and F-measures are used to evaluate the sentiment classification. The result shows that the best lexicon-based technique is VADER with 72% accuracy compared to TextBlob with 70% accuracy. These sentiments results allow for a better understanding and handling of the pandemic. The top three topics are identified and further classified into positive and negative comments. In conclusion, the developed model can assist healthcare workers and policymakers in making the right decisions in the upcoming pandemic outbreaks.
Article
The conventional methods for assessing the performance of educators in academic settings have often been hindered by outdated feedback systems, leading to assessments that can be either skewed or in- complete. This study unveils an innovative, dual-approach methodology that fuses lexicon-driven sentiment analysis with cutting-edge Transformer architectures, specifically focusing on BERT (Bidirectional Encoder Representations from Transformers). This approach provides a more sophisticated, real-time assessment of educators based on student evaluations. One of the standout features of this study is the integration of Speech- to-Text technologies, which allows for the immediate transformation of verbal feedback into text that can be analyzed. The approach makes use of a specialized Educational Sentiment Lexicon for preliminary sentiment evaluation, which subsequently refines the performance of a pre-existing BERT model. This dual-model is proficient in scrutinizing both text-based and verbal feedback—the latter being converted into text through advanced Speech-to-Text techniques. Our results indicate that this approach significantly outperforms existing lexicon-based and machine learning methods in terms of accuracy and comprehensiveness. By providing a real-time, versatile sentiment analysis tool tailored for educational settings, this research marks a paradigm shift in the scope and quality of faculty evaluations, thereby contributing substantially to the field of educational technology and sentiment analysis.
Chapter
Now a days, in the realm of rapid advancement, Artificial Intelligence and Machine learning play a vital role and capture a significant portion of the global market by solving the problems of various sectors like Healthcare, Agriculture, Industry, and so on. This chapter reviews the transformative potential of AIML in the education sector. In this chapter, we observe how AI and ML can enhance personalized learning experiences, streamline administrative tasks, and facilitate data-driven decision-making by exploring various AIML applications and trends including Automated grading, simulation-based learning, etc. Through the exploration, we aim to highlight the significance of incorporating AIML to enhance learning outcomes Additionally, we address the challenges associated with implementing these technologies in educational settings such as privacy concerns and ethical implications. Finally, the chapter offers recommendations for educators and policymakers on utilizing AI and ML to create a more equitable and effective educational system and prepare learners for the future workforce.
Conference Paper
Student feedback and reviews are crucial data that can offer valuable perspectives and insightful observations for enhancing the learning and teaching approach in the classroom. The process of instructors manually analyzing and reviewing numerous feedback to obtain valuable insights is laborious and time-consuming. Sentiment analysis, a subfield of natural language processing, offers the tools to examine, extract, and measure subjective information from textual data. Hence, employing predictive models for sentiment analysis can effectively supplant the laborious job and completely resolve the issue. This work focuses on the implementation of supervised learning methods, ensemble learning methods, and transfer learning techniques for sentiment analysis of student comments. The models employed in various methodologies strive to forecast the sentiment of feedback in an online course as either positive, negative, or neutral. An evaluation was carried out to compare all the machine learning approaches to identify the most precise model with strong performance. The evaluation results indicate that DistilBERT, when used in transfer learning, achieves the highest performance in predicting the sentiment of student feedback, with an F-measure score of 95.49%.
Book
Full-text available
Proceedings of the 4th Workshop for Young Scientists in Computer Science & Software Engineering (CS&SE@SW 2021), Virtual Event, Kryvyi Rih, Ukraine, December 18, 2021
Article
Full-text available
Objetivo: Analisar a percepção da satisfação da experiência acadêmica, dos alunos que participaram de atividades pedagógicas remotas em decorrência da pandemia da Covid-19, comparando-a com o retorno ao presencial. Originalidade/valor: A pesquisa mostra-se relevante na medida em que os resultados indicam a necessidade de mudança na estratégia da oferta de ensino da Instituição de Ensino Superior (IES), que pode resultar na fidelização dos estudantes. Método: A pesquisa conduzida caracteriza-se como descritiva, transversal, com abordagem quantitativa, tendo como participantes 81 alunos de uma IES do Estado de São Paulo. Foi aplicada a Escala de Percepção da Satisfação Acadêmica de Schleich et al. (2006), composta por três dimensões: satisfação com o curso, oportunidade de desenvolvimento e satisfação com a IES. A análise foi desenvolvida com a aplicação do teste de normalidade de Kolmogorov-Smirnov e do teste de diferença entre médias de pares emparelhados de Wilcoxon. Resultados: O estudo demonstra que não há diferença estatística relevante na percepção da satisfação da experiência acadêmica pelos alunos ao comparar o ensino online, ocorrido durante a pandemia da COVID-19, e o presencial, após o término da crise sanitária mundial. Entretanto, a maioria dos alunos (77,8%) desejam uma alteração no formato de ensino ofertado atualmente pela IES, após vivenciarem uma transformação digital (TD), ainda que incipiente, no ensino. Conclusão: O presente estudo contribui com a teoria na medida em que é relatado um fator impulsionador para adoção da TD que não constava na revisão da literatura acerca do tema. A pesquisa aprimora a compreensão sobre as métricas a serem utilizadas pelas organizações, em linha com a lente teórica na qual está embasada, que propõe a utilização de indicadores capazes de capturar o sentimento dos clientes frente à uma TD.
Conference Paper
Full-text available
Online education has grown significantly inpopularity over the last several years, providing students witha flexible and convenient way to study. Despite its manybenefits, this flexible approach of learning requires ongoingassessment and feedback to ensure that it is being improvedand improved. In order to develop a model that can accuratelyclassify learner attitudes as positive, negative, or neutral, thisstudy explores the field of sentiment analysis in the context ofonline course feedback. In this study, we analyzed the students'sentiments regarding their courses based on their feedback ononline courses and tried to predict their sentiments. We haveused the Coursera Course Reviews Dataset, which contains107018 data records of used reviews and comments.Six machine learning techniques (Support Vector Machines(SVM), Naive Bayes, Logistic Regression, Decision Trees,Random Forests, and the AdaBoost) were applied with fourperformance measurement metrics. One of our models,Logistic Regression, outperformed the other models with anaccuracy of 97.31%. Our findings demonstrate that our modelachieved the highest possible accuracy compared to previousrelevant studies in this particular domain.
Article
Full-text available
The COVID-19 has led to the shutdown of educational institutions around the world. Over 1.2 billion students are out of school globally. As a result of COVID-19, the education sector is facing a huge impact on universities forcing their teaching to shift to e-learning and mixed learning modes. During the current emergency, he Learning Management System (LMS) has been a vital resource for virtually all higher education institutions, and a driving force in online learning. After the announcement of University closures, Saudi Arabia is one of the nation where students are worried and going through a phase of learning, where they have no tutor, a lot of course work and only themselves to help them study. Jazan University transforms to E-learning mode of teaching after the suspension of regular classes. The primary focus of this study is to examine the academic semester activities of LMS, Blackboard, amongst faculty members of the Faculty of Arts and Humanities at Jazan University, Kingdom of Saudi Arabia. Furthermore, how the faculty members utilize the learning management system to support their students in the learning process, assessment and include them in the content in hand while they are teaching from home in the midst of the epidemic. The main outcomes of this study are: Role in the achievement of instructive information online, Suitable support provided to the students by faculty and teaching assistants and Contingency strategy for addressing unforeseen online educational system issues.
Article
Full-text available
The initially localized outbreak of Corona transformed into the COVID 19 pandemic, profoundly impacting not just the daily lives and work of people around the world but also the delivery of education to students across the globe due to mass campus closures and suspension of face-to-face teaching. The current study was aimed at identifying how COVID 19 impacted university education in Hong Kong. Data were collected through online structured questionnaires administered via google online forms to students at two prominent universities in Hong Kong. The study revealed that although COVID 19 had negative affected educational delivery, the transition to online modes of teaching by the teachers during the strict lockdown had created a positive image of technology. The study further revealed that having to provide online education had compelled universities with poor infrastructure to improve the means for delivering online learning. In addition to highlighting the damaging impact of COVID-19 on the educational sector, the findings also suggest the need for all educational institutions, educators and learners to adopt technology and to improve their digital skills in line with the emerging global trends and realities in education.
Article
Full-text available
The world has revolutionized and phased into a new era, an era which upholds the true essence of technology and digitalization. As the market has evolved at a staggering scale, it is must to exploit and inherit the advantages and opportunities, it provides. With the advent of web 2.0, considering the scalability and unbounded reach that it provides, it is detrimental for an organization to not to adopt the new techniques in the competitive stakes that this emerging virtual world has set along with its advantages. The transformed and highly intelligent data mining approaches now allow organizations to collect, categorize, and analyze users’ reviews and comments from micro-blogging sites regarding their services and products. This type of analysis makes those organizations capable to assess, what the consumers want, what they disapprove of, and what measures can be taken to sustain and improve the performance of products and services. This study focuses on critical analysis of the literature from year 2012 to 2017 on sentiment analysis by using SVM (support vector machine). SVM is one of the widely used supervised machine learning techniques for text classification. This systematic review will serve the scholars and researchers to analyze the latest work of sentiment analysis with SVM as well as provide them a baseline for future trends and comparisons.
Article
Full-text available
Community's view and feedback have always proved to be the most essential and valuable resource for companies and organizations. With social media being the emerging trend among everyone, it paves way for unprecedented analysis and evaluation of various aspects for which organizations had to rely on unconventional, time consuming and error prone methods earlier. This technique of analysis directly falls under the domain of "sentiment analysis". Sentiment analysis encompasses the vast field of effective classification of user generated text under defined polarities. There are several tools and algorithms available to perform sentiment detection and analysis including supervised machine learning algorithms that perform classification on the target corpus, after getting trained with training data. Lexical techniques which performs classification on the basis of dictionary based annotated corpus and Hybrid tools which are combination of machine learning and lexicon based algorithms. In this paper we have used Support Vector Machine (SVM) for sentiment analysis in Weka. SVM is one of the widely used supervised machine learning algorithms for textual polarity detection. To analyze the performance of SVM, two pre classified datasets of tweets are used and for comparative analysis, three measures are used: Precision, Recall and F-Measure. Results are shown in the form of tables and graphs.
Article
Full-text available
The growth of user’s generated content increased in microblogging platforms like Facebook, Twitter and Blogger in form of client reviews, comments and opinion. Using this bulk of helpful data is difficult to analyze and also a time consuming task. So it is needed to have such an intelligent text mining system that automatically analyze such vast data and categorize them into positive or negative class. Due to the noisiness in data, it is difficult to design such text mining systems because they suffer from mistakes of spelling, grammatical and improper punctuation. Opinion mining is a useful tool to monitor consumer’s feedback and public mood about certain product in terms of negativity or positivity. For example the management of customer relations can use these feedbacks and improve the products by keeping in view the complaints. Lexical tools are one of the famous and useful techniques for sentiment classification. Many extensions and modifications of these tools are available now days. The purpose of this research is to study the available lexical tools and techniques to raise an interest for this research area
Article
Full-text available
Climate change prediction analyses the behaviours of weather for a specific time. Rainfall forecasting is a climate change task where specific features such as humidity and wind will be used to predict rainfall in specific locations. Rainfall prediction can be achieved using classification task under Data Mining. Different techniques lead to different performances depending on rainfall data representation including representation for long term (months) patterns and short-term (daily) patterns. Selecting an appropriate technique for a specific duration of rainfall is a challenging task. This study analyses multiple classifiers such as Naïve Bayes, Support Vector Machine, Decision Tree, Neural Network and Random Forest for rainfall prediction using Malaysian data. The dataset has been collected from multiple stations in Selangor, Malaysia. Several pre-processing tasks have been applied in order to resolve missing values and eliminating noise. The experimental results show that with small training data (10%) from 1581 instances Random Forest correctly classified 1043 instances. This is the strength of an ensemble of trees in Random Forest where a group of classifiers can jointly beat a single classifier.
Article
With the dramatic expansion of information over the internet, users around the world express their opinion daily on the social network such as Facebook and Twitter. Large corporations nowadays invest on analyzing these opinions in order to assess their products or services by knowing the people feedback toward such business. The process of knowing users' opinions toward particular product or services whether positive or negative is called sentiment analysis. Arabic is one of the common languages that have been addressed regarding sentiment analysis. In the literature, several approaches have been proposed for Arabic sentiment analysis and most of these approaches are using machine learning techniques. Machine learning techniques are various and have different performances. Therefore, in this study, we try to identify a simple, but workable approach for Arabic sentiment analysis on Twitter. Hence, this study aims to investigate the machine learning technique in terms of Arabic sentiment analysis on Twitter. Three techniques have been used including Naïve Bayes, Decision Tree (DT) and Support Vector Machine (SVM). In addition, two simple sub-tasks pre-processing have been also used; Term Frequency-Inverse Document Frequency (TF-IDF) and Arabic stemming to get the heaviest weight term as the feature for tweet classification. TF-IDF aims to identify the most frequent words, whereas stemming aims to retrieve the stem of the word by removing the inflectional derivations. The dataset that has been used is Modern Arabic Corpus which consists of Arabic tweets. The performance of classification has been evaluated based on the information retrieval metrics precision, recall, and f-measure. The experimental results have shown that DT has outperformed the other techniques by obtaining 78% of f-measure.