You are on page 1of 16

Text Analysis and Sentimental Analysis on Customer Reviews of

Apple Smart watch SE and Apple Air pod from E-commerce


Website
Abstract
Customer reviews are not only an informational source for businesses to cope with the
competition, but they also play an important part in providing information to customers looking
to buy a product. Customer reviews of the Apple Smart watch Se and Apple Air pod were
evaluated in this research report from two major e-commerce websites: Flipkart and Amazon.
RStudio is used for text analytics and Lexicon-based sentiment analysis of customer evaluations.
Text mining has also been used to examine customer feelings and emotions. The goal of this
paper is to assist corporations in determining whether or not their products are accepted by
customers.

KEYWORDS – Text Analytics, Sentiment analysis, online reviews


INTRODUCTION
Advancement in technology has improvised many aspects such as communication and freedom
of speech in various domains such as social networking sites, blogs, and other websites. Social
media is one such platform that has completely transformed our lives. Through technology, we
can remain connected with one another and the general environment (Coleman, L. J., Chandler,
K., & Gu, J., 2013). In this modern era of the 21st century, several different forms of text prevail
such as articles, social media posts (Twitter, Facebook, Bloomberg), news reports, e-mails, and
other additional reviews on various websites. It is cumbersome and difficult to search the trends
of the vast and massive information available on the internet. The text mining approach can be
used to find out some meaningful results among these gigantic datasets. Text mining is a highly
used technique used by researchers to find out the results, trends and analyze the raw data. Text
analytics can be done through various software such as Semantria for Excel, Meaning cloud,
Parallel dots, CX data science add-ins in Excel. Sentiment analysis can also be done by
WORDSTAT, NVIVO, STATISTICA. For the current study, we have employed numerous
packages (tm) (wordcloud) (wordcloud2) (pdftools) (qdap) (meanr) (SentimentAnalysis) using
RStudio software (Version 1.4.1106). The study provides the analysis of the customer-generated
content by anatomizing product reviews dropped on the e-commerce websites by experienced
customers.
Following the introduction section, extant literature has been discussed explaining the use of text
analytics and sentiment analysis. The remainder of the paper advances research methodology,
which explains the detailed process of text analytics and sentiment analysis. The last section
includes the discussions and conclusions of the study conducted.
E-commerce
In today's commercial world, e-commerce is booming. Electronic commerce is referred to as e-
commerce. E-commerce (Electronic commerce) entails the purchase and sale of goods and
services, as well as the transmission of payments and data, over an electronic network, most
commonly the Internet. Electronic commerce (E-commerce) is a paradigm change that affects
both marketers and customers. E-commerce, on the other hand, is more than just another
technique to improve existing company operations. It is pioneering a complete transformation of
the traditional business model. This huge shift in business paradigm is gaining traction all across
the world, and India is no exception. Environmental consequences Although the concept is
widely employed in today's commercial environment, it has yet to be fully explored.
Flipkart

Flipkart was created in 2007 by Sachin and Binni Bansal, IIT Delhi students and former
Amazon.com employees. Flipkart is a Singapore-based e-commerce company that operates in
India. Flipkart is one of India's most popular websites, according to alexia internet.
WS retail, a subsidiary of Flipkart, sells goods in India. Other third-party dealers or businesses
can also sell goods on Flipkart's marketplace.
Flipkart began selling books in 2008, but as the company grew, it expanded its offerings to
include consumer electronics, apparel, home décor, appliances, cosmetics and fashion products,
and more. Flipkart has risen to the top of the Indian market because to its extensive network and
efficient customer connection management. Flipkart accepts cash on delivery, net banking, debit
or credit card transactions, e-gift vouchers, and card swipe on delivery as payment methods.
Amazon

By total sell and market capitalization, Amazon is the world's largest online marketplace. Fast
start bookstore, eventually broadening to sell DVD, VIDEO, CD, SOFTWARE, GAMES, MP3
and consumer goods, food, and furnishings. Amazon.com, which was created by Jeff Bezos in
1994 and is headquartered in Seattle, Washington, is an American electronic commerce firm. On
July 5, 1994, Jeff Bezos incorporated the firm as "Cadabra," and the site went live as
Amazon.com in 1995. Because cadabra.com sounded too much like cadaver, Bezos changed it to
amazon.com. Furthermore, a name beginning with "A" was preferred because it was more likely
to appear at the top of any alphabetized list. It is the United States' largest Internet-based
enterprise. Amazon.com began as an online bookstore, but has now expanded to include DVDs,
VHS tapes, CDs, video and MP3Downloads/Streaming, Software, Video Games, Electronics,
Apparel, Furniture, Food, Toys, and Jewellery. The company also makes consumer products, like
as the Kindle, Fire Tablets, Fire TV, and Phone, and is a significant cloud computing service
provider.
Product under the Study
a) Apple smart watch SE

Thanks to its unique features, the Apple Watch SE will help you stay active, connected, and
more. Its fall detection feature will assist you in obtaining medical attention in the event of a fall.
To do so, simply press and hold the side button. It reminds you to wash your hands, keeps track
of your menstrual cycle, and so much more. You may also use Siri to quickly look up nearby
eateries, song titles, and other information.

b) Apple Air pod

The all-new and improved Apple Airpods Pro with Charging Case is here, with a design and
capabilities that will improve your hearing experience. The AirPods Pro link to your iPhone or
Apple Watch in a remarkable way. Your favourite personal assistant is summoned with a simple
"Hey Siri" command. Without lifting a finger, you can control your music, calls, volume,
directions, and more.
Literature Review

Arwa S. M. AlQahtani: Sentiment analysis is a necessary and widely used method for
collecting information from text data on E-Commerce websites. In the form of ideas, feedback,
tweets, and comments, e-commerce portals generate a vast volume of text data every day.
Furthermore, reviews, ratings, and emoticons imply the public's opinion. Extracting product
information from a review will assist a customer in learning more about the product and making
a decision.

Conor Gallagher: Businesses must have a clear line of sight into what their consumers believe
about the company, its goods, people, and how it treats them in today's environment of ever-
growing customer data. The findings show that a company cannot rely on a single statistic as the
only source of truth when it comes to customer experience. There is a considerable disparity
between the customer rating score and the mood expressed in their product review. The authors
argue that a company's customer feedback scores must be supplemented by a strong sentiment
analysis strategy.

Han, Hyun Jeong According to an analysis of the text of 5,830 reviews spanning 57 hotels in
Moscow, Russia, hotel ratings do not give the whole picture about how visitors perceive a hotel.
Negative comments, for example, carry a higher weight in a guest's hotel rating than good
comments, according to the study.
Research Design

Yoosin Kim: Text mining and sentiment analysis have been widely used in assessing electronic
word-of-mouth since the advent of internet communications. The goal of this research is to create
a domain-specific vocabulary of sentiment analysis to forecast box office performance in the
Korean film market and to test the lexicon's viability. The findings revealed a stronger positive
association between box office success and customer sentiment, as well as a substantial positive
influence in the prediction model's linear regression. Furthermore, it demonstrates that emotion
in user-generated material can be a better predictor of business success.

Objective of the study

1. To analyze the customer reviews of apple smart watch se from the e-commerce websites
2. To find the difference and similarity between various frequently used words from the
customer reviews
3. To analyze the sentiment of customer reviews

RESEARCH METHODOLOGY
The following methodology is used in this study to better comprehend customer dialogues and
discussions on virtual shopping platforms. There has been an increase in online purchasing and
conversation on e-commerce platforms in recent months. The focus is on social media, with
conversations centred on the product's performance and user experience. As a case study for this
research, a comparison of the two largest e-commerce websites, Flipkart and Amazon, is
undertaken. The choice of these companies was based on their significant market share. Flipkart
has a market share of roughly 31.9 percent, whereas Amazon has a market share of
approximately 31.2 percent. Customer reviews of the Apple Smart Watch SE and Apple Air pod
were subjected to sentiment analysis and text analytics, and the customer evaluations were
copied from the websites indicated above.

Type of Study – Descriptive


Sample- customer reviews of product under study from flipkart and amazon
Size- A total of 600 customer reviews
Software- R-studio and Python
Type– Secondary data
Techniques used in study- Text analysis, Sentimental analysis, Word cloud

TEXT MINING
Text mining includes many things such as -Data Collection, Data Pre-processing, and Data
Transformation, Data Analyze and Result evaluation Text mining techniques are essential for
handling unstructured text data. Generally, the procedure of text mining involves the process of
constructing input text, such as adding derived language features, parsing and removing
unnecessary characters. A variety of technologies have been studied to summarize and
understand the data required to obtain business insight from the rapid growth of social big data
such as blogs, the web, and Twitter [4-7]. Therefore, to grasp trend or obtain insight from social
big data, it is necessary to group frequent words obtained through text mining on the basis of
association and to integrate them by topic. However, a web document composed of several
sentences such as a blog may include two or more subjects in a document, and a tweet sentence
composed of a short sentence due to the length limitation data can be extracted in a small amount
of information from the text. Therefore, it is difficult to grasp the contextual meaning of
keywords by extracting nouns or adjectives included in the text and deducing the entire contents
depends upon the appearance rate of repetition of the words. So, in the current study we have
tried the information from the various websites to get adopt the text mining process in an
effective manner. This study proposes a topic-oriented analysis method consisting of related
word group through text clustering which improves data analysis method based on word
frequency.

Process of Text Mining


Text mining is the procedure of changing unshaped data to shaped data for drawn out important
information. Figure 1 shows the whole process of text mining. There are five steps under text
mining operation: Data Collection, Data Pre-processing, and Data Transformation, Data Analyse
and Result evaluation.

a) Data Collection
Under this step data in unstructured form is being collected from several sources and it
can be taken in the form of blog, reports, news and reviews.

b) Data Pre-processing
In this step the repetition, separate words, stemming and inconsistencies are being
removed as gathered data was preprocessed for the same. In the remembrance, the data
was divided into single word i.e. token.
c) The information contains unwelcomed words like, an,a, the, be that as it may, and, of,
and so on. These words are called as stop words. Stop words are evacuated in this
progression.
A stem is a characteristic gathering of words with equivalent (or fundamentally the same
as) which means. This technique portrays the base of specific word. Inflectional and
derivational stemming are two sorts of strategy. One of the well-known calculations for
stemming is watchman's calculation. for example, on the off chance that a record relates
word like acquiescence, surrendered, leaves then it will be considered as leave
subsequent to applying stemming strategy.
d) Data Transformation
Information change intends to change over content report into the pack of words or vector
space archive model documentation, which can be utilized for additional compelling
investigation.

In include extraction, the valuable significance words are extraction from the record. In
highlight choice, pertinent words are chosen. There are two techniques in include choice
for example separating and wrapping strategies.
e) Data Analysis
The handled information was broken down utilizing content mining strategies, for
example, data recovery, order, characterization and rundown.

Data handled in the above advances is utilized to extricate significant and applicable data
for powerful and convenient dynamic and pattern examination.
f) Evaluation
This progression assesses the outcomes as far as exactness, review and precision.

Detail Steps to be involved in text mining

Given underneath rundown of steps further shows the arrangement of capacities which
are acted so as to make the five-advance based procedure increasingly powerful for
content mining in R package
Read file using readLines

Split line to words based on " "

Unlist single vector of words

Clean vector of words

Remove numbers/digits

Remove punctuations

Remove special chractors

Remove white space

Remove empty vectors

Convert vector of words to dataframe

Clean Dataa Frame

Remove Stop words

Remove Sparse words

Convert vector of words to dataframe

Perform operations as requiered

Plot graphs as required


Text Mining Techniques:
The different utilized in the content mining are talked about beneath:
a) Clustering
clustering is a solo procedure to characterize the content records in bunches by applying
distinctive grouping calculations. In a bunch, comparative terms or examples are gathered
extricated from different reports. Bunching is acted in top-down and base up way. In
NLP, different sorts of mining apparatuses and procedures are applied for the
investigation on unstructured content. Various methods of grouping are progressive,
dispersion, thickness, centroid, and k-mean [8]. Zhang et al. [9] utilized cosine to
ascertain a connection likeness between two anticipated reports in a low-dimensional
semantic space and performed record bunching in the relationship closeness measure
space.
b) Categorization
Content arrangement is otherwise called content characterization, or theme spotting is the
errand of naturally arranging a lot of archives into classifications (or classes, or points)
from a predefined set [10].
c) Summarization
Content synopsis is a procedure of gathering and creating compact portrayal of unique
content archives [11]. In past programmed content synopsis was performed based on
event a specific word or expression in report. Later on, extra strategies for content mining
were presented with standard content mining procedure to improve the pertinence and
precision of results [12]
d) Sentiment Analysis
Supposition examination is a sort of characteristic language handling for following the
temperament of people in general about a specific item or point. Assumption
examination, which is likewise called conclusion mining, includes in building a
framework to gather and inspect sentiments about the item made in blog entries, remarks,
audits or tweets. Supposition investigation (or sentiment mining) is characterized as the
undertaking of finding the assessments of creators about explicit elements. The dynamic
procedure of individuals is influenced by the feelings framed by thought pioneers and
conventional individuals. At the point when an individual needs to purchase an item on
the web, the person in question will normally begin via looking for surveys and feelings
composed by others on the different contributions. Slant investigation is one of the
sultriest research regions in sociology. In this way, in the preset investigation likewise we
have utilized the slant examination to know different assessment of different profound
masters through the recurrence of keys for the most part utilized by them and which draw
in the individuals toward a specific master and the foundation.

According to process of social media mining, twitter sentiment also has gather, analyze and
visualize stage.

Data Collection
All reviews of Flipkart and Amazon were fetched from the websites, and 150 reviews of Amazon
and Flipkart were collected for Apple Smart Watch and Apple Air Pod. A total of approximately
600 reviews in total were collected, and the first review comment was posted on September
2019, and the latest one was on 7 April 2021.
Once the comments were copied into a text file, then pre-processing of data was done. The
comments only in the English language were included in the current study. As per the
requirement of text analytics full document was converted into lower case. Post that, the text was
cleaned because the packages have some essential prerequisites for the processing analysis of the
text. The data was cleaned by removing numerical data, punctuation marks, special characters,
and emoticons
FINDING AND RESULTS
Word cloud and Wordcloud2
Wordcloud elucidates the most frequent words used in the text file. In the current study,
Wordcloud expounds the most common terms used by the customers while expressing their
opinion through their reviews on the website. Following the above process, wordclouds were
generated using the tm package. The most pertinent characteristic of wordcloud is that it
highlights the most frequently used words, which eventually spills the experienced customers
ideas about that particular product. Initially, the data was cleaned, the corpus was created, and
the Wordclouds were generated from the same.

Wordcloud of all files


makes
lovecomes
cellular version

user
pods replacement

replace
return
airpod
connectivity devicesfeaturesince
screen chargingtransparency fast

got serialsale
first want
hear deliveryconnect

mode
getservice every

used
right
w arranty don’t thing
android
know
say music better bass days perfect
cancellationworthbluetooth
ecosystem

definitely
also

tried
amazing

instead
justbatteryiphoneaudio lookvoice

still
must
however
buying getting

store
canusedont

apple

check purchase alwayscharge


lifelow

works
anc
good
experience

sound
back looking

bigfit

it’s time
without
reviews wirelessawesome
sony

calls
product
ear value
i’m

watch
display give

one
prolike
greatwell
overall keep
listening months

working

airpods
now long
flipkart new
last day ipad
around

ears

earbuds pretty
genuine

quality
number

expectthink issue
seriesusing

headphones
hours

noisebit left will


bad
case spatial
much

center really
buy

thoughwork
designoriginal
buds care amazon
phone
feel even
call
ecg

best

wear
many

smart
high
earphones device ios pricefeatures
people
box

need lot
little

madereceived
makeproductsboughtmoney
found

comfortable
pair

backup take
active endfakebose customerplease
said

compared
able

supportw rist
different issues problem
defectivesure never
something trackingside
Wordcloud 2 of all files
Comparison Cloud
The Comparison Cloud describes and marks out the relevant information from raw data. It is one
of the most efficacious description of comparative analysis of the different sets of data. The
following figures explain the differences in the opinions expressed by the customers for the same
product

reading came
extra perfect w ell support
doesnt

tried
A_Watch A_Airpod
notifications bought tipspods
found
sleepbad new
version
get buy always got
budget w₹
atches replacement
makes
series

displayproducts

wireless
price
size
latest connected

spo day can said issue


screen
smartw atch lifewiv eillnumber volume hours
serviceearphones
purchase

give
watch
make it’s music
festival

charge

w arranty
device
days

appleamazon
heart

tracking wrist
less

box ears customer


earbuds
serial

gps ear pair differentupdate


let

airpods care

center
lasts look still
sound may

mustindianaccuracy
used

replace asked
fake

lot
ty peblooddont received
product

much
bass
useecg

upgraded
noise
times
feelbose thanksback small
many
productawesome
time one
totally

experience anc
ittill

data
using
flipkart
mode

fit money
activebest hear
headphones
market

value
spatial

wordssmart

ios

quality
loving

track
airpodcase justbatteryfull
upgrade

beastnow
applei
audio

procellulargood
great

say devices right car


jabra like talkwant build
love
works amazing call lovers
workthing
cancellation awesome

F_Airpod F_Watch
hrs
yetit except
worth

doubti
connectivity

strap without
don’t

transparency
sony features rate productgreat
left

delivery

buds level
stands
cycling
watching

also first calls voice


postpaid

tiny comfort seamless working


backup

fitness rest
phone

better
option

packaging expected excellent


done

feels top convenient definitely


nice
really w ish
app itperfect
amazinggreat
return
Sentiment analysis
Sentiment analysis is the most conventional method used to identify the reactions and emotions
of the customers. It helps to explain what the customer feels after purchasing the product and
their experience regarding the same. Sentiment analysis helps to recognize the emotions by
classifying them into positive, negative and neutral. The further classifications were done into
different categories which expel emotions like anger, fear, disgust, anticipation, joy, sadness,
surprise, trust. All the above emotions were classified using the Lexicon-Based Dictionary.
Different emotion category is built up for different words such as for anger, fear, disgust
(negative emotion) and the other emotion category may include happy, joy (positive emotion).
The nrc and syuzhet package were used to conduct the sentiment analysis.
The current study employs the sentiment analysis to recognize different emotions expressed by
the users of all the four e-commerce companies.
The following process was employed to find the emotions expressed by the customers.
Firstly, the function readlines was used to read the text in order to determine the emotions of the
customers expression. UTF-8 encoding is by default but here we choose ASCII. Then the text
was cleaned because the packages have some essential prerequisites for the processing analysis
of the text. The data was cleaned by removing numerical data, punctuation marks, special
characters, and emoticons. Further processing is also required to plot the graphs of Sentiment
analysis.

You might also like