A Company’s Corporate Reputation through the Eyes of Employees

Measured with Sentiment Analysis of Online Reviews

R. E. Loke

and R. Lam-Lion

Centre for Market Insights, Amsterdam University of Applied Sciences, Amsterdam, The Netherlands

Keywords: Sentiment Analysis, Corporate Reputation, Natural Language Processing, Semantic Search, Scraping.

Abstract: Corporate reputation can be defined as the overall assessment of a company’s performance over time (Kircova

& Esen, 2018). Organizations with a positive corporate reputation create a competitive advantage and are

more likely to influence customer’s behaviors and attitudes (Kircova, 2018). Measuring corporate reputation

from online data is an increasingly important area in business studies because the amount of opinions and

comments is increasingly growing on the internet and has become very accessible to strangers (Shayaa, 2018).

Traditionally, corporate reputation is measured with well-known approaches such as surveys, qualitative

interviews, and sample groups (Smith, 2010). Researchers like Fombrun, Fonzy and Newburry (2015)

developed instruments to measure corporate reputation and predictivily modeled its impact on stakeholder

outcomes. So far, however, there has been little attention in the literature on sophisticated measurement

techniques for corporate reputation that can be applied to online reviews from the public web. This paper

applies sentiment analysis in combination with semantic search as a suitable technique to explore how

employees perceive organizations. By using our toolbox, organizations can adapt to market changes and cater

to stakeholders’ needs. Also, it can be used to raise awareness for organizations that are unaware of negative

reviews online.

1 INTRODUCTION

Sentiment analysis is without any doubt a useful

means to measure the public's opinion and an

organization that gathers information about its

stakeholders that is tagged with sentiment scores is

doing this to be able to understand public opinion and

to improve the quality of its products (Shayaa, 2018).

In this paper, we specifically focus on the

employees' opinion in public opinion as this

represents a relevant stakeholder for many

organizations.

This paper's main research question is as follows:

"How can corporate reputation from the perspective

of the employee be measured from online reviews

with the help of sentiment analysis?" The question

will be answered with both primary research in which

we propose a toolbox for measuring the corporate

reputation construct from online employee reviews

on the popular public website nl.indeed.com, in

section 3, and literature study, in section 2. Section 4

concludes the paper.

https://orcid.org/0000-0002-7168-090X

2 LITERATURE REVIEW

Many studies have dedicated methods to measure

corporate reputation (Fombrun, 1996). Most of them

have conducted these studies through qualitative

interviews, surveys, and focus groups (Smith, 2010).

From a socio-economical perspective, the importance

of measuring corporate reputation is crucial. This

means that to prevent damaging a brand’s image,

analytical insights from this study should help assess

the problem that organizations fail to recognize poor

corporate reputation (Lange, 2011). If organizations

detect negative reviews from an early stage, they can

steer a positive message and negative communication

exiting the organization.

2.1 Impact of Corporate Reputation

Many researchers define corporate reputation.

According to Bronn, reputation has many

interpretations, but most agree that it is an intangible

asset that can build a competitive advantage for a ﬁrm

378

Loke, R. and Lam-Lion, R.

A Company’s Corporate Reputation through the Eyes of Employees Measured with Sentiment Analysis of Online Reviews.

DOI: 10.5220/0010620603780385

In Proceedings of the 10th International Conference on Data Science, Technology and Applications (DATA 2021), pages 378-385

ISBN: 978-989-758-521-0

(Bronn, 2015). A good reputation can give an

organization a competitive advantage (Bronn, 2015).

This is because a good reputation is valuable, rare,

and cannot be replaced (Bronn, 2015). Other

researchers argue that corporate reputation is defined

as a stakeholder's overall assessment of a company's

performance over time (Kircova, 2018). It reflects

multiple stakeholders’ perceptions about the

organization's effectiveness (Kircova, 2018).

Companies with a high reputation create a

competitive advantage and are more likely to

influence customers' behaviors and attitudes

(Kircova, 2018).

According to Smitha, brand associations affect

the corporate image and is the company's reputation.

A positive brand image can boost consumer

confidence in a company, affect decision making, and

build loyalty. Companies take great care in managing

their brand image because it is a source of value

(Smitha, Smith & Wang 2010). In general, there is an

agreement that its stakeholders determine corporate

reputation and that it is an intangible asset that creates

a competitive advantage and is of high value to an

organization. Since it is seen as an asset, damaging

the organization with negative reviews might

severely damage the organization.

2.2 Stakeholder Tracking and

Analysis: The RepTrak System for

Measuring Corporate Reputation

The RepTrak System describes the importance of

using a prototype system that was developed to

measure corporate reputation. It is an analytical tool

used to track stakeholders’ perceptions of companies

(Fombrun, 2015). A seven-dimensional framework is

described, derived from prior literature that highlights

how stakeholders should be managed (Fombrun,

2015). It can be used as a basis to verify which

dimensions are touched upon in online reviews that

are gathered from the public web. The seven

dimensions are described in the subsequent

paragraphes.

First, Products and Services refer to the perception

that reputation is likely to be affected by the products

that an organization offers. A company is usually

positively perceived if it offers high-quality products

and services (Fombrun, 2015). According to Smith

& Wang, a company’s brand name not only indicates

the type and quality of products and services offered

but also symbolizes the character of the company

(Smith, Smith & Wang, 2010). Consumer perception

of a brand can make or break a company (Smith,

2010). A positive brand image can boost consumer

conﬁdence in a company, affect decision making, and

build loyalty. Online reviews about a product or

service are likely to be put online by customers

(Smith, 2010). In this particular study, customers

were not investigated. Even though customers are not

investigated, it is verified whether employees include

opinions about the focal organization’s products and

services in online reviews.

Second, Innovation indicates that an innovative

organization adds to a positive company reputation

and strives to do new things within the organization

(Fombrun, 2015). According to Courtright, there are

two basic views on innovation. From an

organization’s point of view, innovation concerns its

ability to create new and better products and services

(Courtright & Smudde, 2009). From a market’s point

of view, innovation is about how it introduces new

and better things into the socio-economic systems in

which organizations take part. By stating this is meant

that innovation can involve employees in the process

and could be a topic, they write about in online

reviews. The study from Courtright & Smudde

suggests that employees receive multiple messages

within and outside the organization and that good

corporate communication can prevent negativity

exiting the company, even if this is not necessarily

about innovation (Courtright, 2009).

Third, Workplace suggests that most stakeholders

like and respect companies that maintain healthy

workplaces and do good for their employees, which

is also visible outside the organization (Fombrun,

2015). According to Nolan, an employer brand image

refers to the package of functional, economic, and

psychological beneﬁts provided by employment, and

identiﬁed with the employing company (Nolan,

2013). It refers to a person’s beliefs about what it

would be like to work for a corporation (Nolan,

2013). Fombrun suggests that most stakeholders like

and respect companies that maintain good workplaces

(Fombrun, 2015). Both studies indicate that the

workplace can impact the brand’s image and can be

affected if viewed negatively. Besides, Nolan states

that satisfied employees are more likely to commit to

long-term involvement and act as ambassadors of the

company and give an employer a favorable rating

(Nolan, 2013). While being aware of this, it is

especially crucial when analyzing online reviews

gathered for this research.

Fourth, Governance is considered a key part of

reputation management in terms of being transparent

and ethical; organizations with corporate governance

tend to build more trust with stakeholders (Fombrun,

2015). According to Davis, corporate governance can

be defined as the structures, processes, and

A Company’s Corporate Reputation through the Eyes of Employees Measured with Sentiment Analysis of Online Reviews

379

institutions within and around organizations that

allocate power and resource control among

participants (Davis, 2005). It could be that these

structures, processes, and institutions are also

discussed in online reviews.

Fifth, Citizenship is an essential aspect of

reputation management since companies are admired

for good deeds and building a strategic asset if they

participate in, for example, a good cause (Fombrun,

2015). A qualitative study suggests that stakeholders

tend to respect and admire a company for their good

deeds. For example, good deeds such as health

benefits, care plans, and other benefits could lead to a

positive corporate review (Orlitzky, 2012).

Sixth, Leadership is a visible quality that aids in a

good reputation. A good leader attracts positive

media coverage and is more likely to support

company activities. Fombrun’s study mentions that a

good manager could influence employees to have a

positive work environment and stimulate employees

to bring harmony to the workplace (Fombrun, 2015).

Good leadership is very likely to be discussed in

reviews since it directly affects employees (Fombrun,

2015).

Seventh, Performance is seen as a signal that

influences how stakeholders assess companies. A

strong financial performance is positively attributing

to an organization’s trustworthiness (Fombrun,

2015).

These seven dimensions affect corporate

reputation and are necessary to manage and satisfy

stakeholders.

2.3 Risk Governance, Structures,

Culture, and Behavior: A View

from the Inside

A different approach for measuring corporate

reputation is to conduct surveys. This study suggests

that internal corporate governance and reputation

could be measured with surveys held from

employees, operating in different business units

(Sheedy & Griffin 2018). It was seen that employees

generally provide favorable ratings for training

programs, training frameworks, and managers

(Sheedy, 2018). In contrast, ratings for remuneration

were less favorable (Sheedy, 2018). One of the

findings was that leaders play a crucial role in

developing culture (Sheedy, 2018). However, the

paper indicates that biases occurred because of a

higher proportion of males in the surveys. Even

though this paper researched the banking sector,

surveys were seen as a convenient yet obsolete

method to measure reputation.

In our approach, the use of corporate reviews

eliminates the bias of managers influencing

employees to give favorable answers. Nevertheless,

Sheedy’s paper indicates that it was challenging to

obtain survey access and that a self-selection bias

could be present.

2.4 Sentiment Analysis of Big Data:

Methods, Applications, and Open

Challenges

Shayaa, Jaafar, Bahri, Sulaiman, Seuk Wai Chung,

Piprani & Al-Garadi (2018) highlight the

development of sentiment techniques and discuss

non-technical challenges on how sentiment analysis

is applied. The amount of opinions and comments is

increasingly growing on the internet and has become

accessible to strangers (Shayaa, 2018). Since large

amounts of information could be gathered from the

internet, text mining is one of the approaches for

analyzing text (Shayaa, 2018). Text mining uses

Natural Language Processing, knowledge

management, data mining, and machine learning

techniques to process text documents (Shayaa, 2018).

Text mining/analytics are originally conducted for

two purposes. The ﬁrst purpose is to analyze people’s

sentiment on an issue or phenomenon (Shayaa, 2018).

Hence, sentiment analysis goes through massive

amounts of data to identify people’s attitudes,

thoughts, judgments, and emotions on an issue

(Shayaa, 2018).

The second purpose is to assess people’s opinion

on a product, person, event, organization, or topic

from a user or group of user perspectives. Similar to

sentiment analysis, opinion mining is a natural

language processing task that uses an algorithmic

technique to recognize content with opinions and

allocate it to positive, negative, or neutral polarity

(Sheedy, 2018).

These studies highlight that an organization that

gathers information about their stakeholders and

applies sentiment analysis is doing this to improve the

quality of their products. However, it is indicated that

additional factors might play an active role in better

understanding public opinion, or employees’ opinion.

These factors might include competitors and

economic growth. This illustrates that sentiment

analysis is a useful means of measuring public

opinion and that organizations can recognize public

opinion more efficiently with sentiment-driven

applications.

DATA 2021 - 10th International Conference on Data Science, Technology and Applications

380

2.5 Journal of Business Research: The

Role of Emotions and Conﬂicting

Online Reviews on Consumers'

Purchase Intentions

A recent study of online reviews demonstrates the

importance of online review content. It stresses the

relevance of encouraging consumers to write more

elaborate, realistic, and informative comments instead

of using the platform to express emotions (Ruiz-Mafea,

2018). In this way, online reviews increase in quality

and perceived helpfulness (Ruiz-Mafea, 2018). It is

also meant to highlight that most reviews tend to be

non-informative and could have fake content, leading

to potentially harming the organization. The study also

demonstrates that users that provide text and pictures

are seen as more accurate and useful online reviews

and leads to quicker decision-making of consumers

(Ruiz-Mafea, 2018). The following six factors were

considered to be important in determing online

review’s quality: (1) Online review credibility: the

overall opinion of what consumers thought about the

review, whether the review was factual, accurate and

credible; (2) Online review informativeness: this

concerned the relevance, completeness, and timeliness

of the review; (3) Online review helpfulness: whether

consumers were not raised with any more questions

after seeing the online review; (4) Online review

persuasiveness: whether consumers were convinced

with arguments after seeing the online review; (5)

Empathy: the extent to which consumers felt

concerned or moved by the review; (6) Emotions: these

emotions were allocated into various categories such as

angry/content, displeased/please, bored/entertained etc

(Ruiz-Mafea, 2018).

This study helps understand the type of reviews

that generate a positive decision-making process; this

also applies to employees that work for an

organization. Besides, according to a Bright Local

Survey, 82% of potential employees read online

reviews of the companies on employer review sites

(Murphy, 2019). For this reason, it can be assumed

that it affects corporate reputation. For example,

according to Muse, it should be considered that a

dissatisfied employee is more likely to post a negative

review and express his frustration than a satisfied

employee (Wolf, 2019). A satisfied employee is

likely to tell three close friends about a positive

experience in a company, whereas a dissatisfied

employee could feel the need to let more people know

about a bad experience in an organization (Wolf,

2019). For this reason, it could imply that negative

reviews that are posted are usually intended to harm

the organization (Wolf, 2019).

3 RESEARCH METHODOLOGY

The application that was developed for this research

helps in recognizing negative reviews and can raise

alarms for organizations. By offering an application

that can perform sentiment analysis on online

reviews, organizations will not fail to adapt to market

changes. This section discusses the research

approach, primary data collection as well as data

analysis applied.

3.1 Research Approach

This study applies text analysis. Text analysis is a

research method that is being practiced to identify

patterns of texts and uses both qualitative and

quantitative research.

Qualitative research involves gathering and

analyzing non-numerical data, understanding

concepts, opinions, or experiences (Bhandari, 2020).

An example of qualitative content analysis in this

study is the exploration of Fombrun’s dimensions in

the review text. The words associated with these

dimensions are a preliminary means to understand

what employees are writing about whereafter

sentiment analysis can be performed.

Quantitative research involves collecting and

analyzing numerical data for statistical analysis

(Bhandari, 2020). An example of quantitative

research in this study is calculating the frequency of

words used in a review, the sentiment of reviews from

a particular organization, and the collection process

of online reviews.

Since there is no existing literature on measuring

corporate reputation with online reviews, this study

intends to be exploratory.

The application should give directions to

organizations, yet it could be developed to a further,

more technical extent.

3.2 Primary Data Collection

In order to measure corporate reputation, primary

data has been collected through the process of web

scraping. Web scraping can be defined as the

construction to download, parse, and organize data

from a website in an automated manner (Broucke &

Baesens, 2018, pp 1-2). Web scraping is considered

useful, especially for data scientists, because it

provides a raw form of information (Broucke, 2018).

However, structuring the data can be challenging

(Broucke, 2018). Many data scientists use API's to

scrape data (Broucke, 2018). An API is an

Application Programming Interface that can be used

A Company’s Corporate Reputation through the Eyes of Employees Measured with Sentiment Analysis of Online Reviews

381

as an instrument to access data in a more structured

manner (Broucke, 2018). According to Indeed, the

organization reserves the right to place limits on

access to any API (without limitation on number of

calls or requests) and monitor the usage to enforce

these limits (Indeed, 2020). This means that websites

such as Indeed do not provide API for free. In order

to obtain an API from Indeed, it requires to become

an editor. For that reason, web scraping has been

utilized.

Indeed has been chosen as a target website

because it is a popular international website, and there

are a large number of corporate reviews available.

Other websites, such as Glassdoor.com, were not

chosen by us due to the scope of our research design.

The collected data can be considered as primary

because it is collected for the sole purpose of this

study. For instance, many researchers use scraped

data to develop deep learning models, natural

language processing, and competitor analyses

(Broucke, 2018).

Similarly, natural language processing is used for

this research. It can be defined as the technology that

aids computers in understanding human language

(Garbade, 2018).

Before commencing the study, ethical clearance

was sought because web scraping could be unethical,

even illegal, if used for commercial purposes.

Moreover, the web scraping process started with

analyzing the reviews beforehand. This process was

completed by inspecting the HTML page and

developing a spider, a Python code that extracts data

from Indeed.nl.

For the web scraping process, Scrapy, a Python

framework, was used. This framework is utilized for

large scale web scraping and is frequently used by

engineers because of its simplicity and easy access

(Rizvi, 2017). Just as Scrapy, there are other scraping

tools available such as Lxml, BeautifulSoup,

MechnicalSoup, Python Requests, and others.

Among these, Scrapy and Beautiful Soup tend to be

popular among developers (Choudhury, 2020).

The reason for working with Scrapy is because the

researcher was familiar with Scrapy before starting

this project.

A few advantages of using Scrapy are: (1) Open-

source framework. (2) An interactive shell console

allows trying out the CSS and Xpath expressions to

scrape data. If the right information extracts from the

web page, the selectors are added to the spider. (3)

Prior knowledge to use Scrapy was present. (4) Large

community of developers.

It is prevalent by developers to use Scrapy in a

Conda environment because of its ease-of-use.

However, for this particular project, Docker was used

to ensure a well-functioning portable virtual

environment. Using Docker, the researcher

guaranteed that the software would behave in the

same way, regardless of where it is developed

(Anderson, 2016). This was done by creating multiple

Docker containers.

As mentioned, the prototype that was built is a

web crawler that extracts data from Indeed.nl. This

prototype is needed to collect data.

The collection of data consisted of 893 reviews.

Using Pandas, a Python library, the JSON data

format was converted into a CSV. A python code

enabled converting the data format because using a

CSV format was preferred.

The reviews consisted of the following variables:

title, author, review text, date, and rating. These

variables were all variables that were available and

were assigned to structured data columns.

In total, online reviews from seven organizations

were scraped (see Figure 1); these companies were

randomly selected. The organizations do not operate

in a similar sector. The online reviews were from

DHL (291 reviews), Coolblue (96 reviews),

McDonald's (150 reviews), Bol.com (37 reviews),

Nike (78 reviews), Ikea (127 reviews), and Shell (114

reviews).

The reviews were collected during April 2020.

Figure 1: Number of reviews from indeed per organization.

3.3 Data Analysis

The first step in analyzing the data was to store the

data in a relational database. Storing data in a

database can be seen as vital because it can manage

large amounts of data. If it is decided to reproduce this

study on a larger scale, local file storage is not

recommended (BBC, 2020). This step is necessary to

ensure that the reviews were accessible from a

database.

Just as for using Scrapy, a Docker container was

created to work with PostgreSQL and PGadmin.

Since the dataset was relatively small, the data is

managed with a local CSV file. After storing the data

in a relational database, queries were created to find

Fombrun's dimensions in the corporate reviews

(Fombrun, 2015) with the following words that are

DATA 2021 - 10th International Conference on Data Science, Technology and Applications

382

associated with Fombrun's seven dimensions

(Fombrun, 2015): (1) Workplace: work, atmosphere,

colleagues, fun, over time, building, and environment.

(2) Governance: ethics, open, transparent, fair in

conducting business. (3) Products & Services:

delivery, product, and service. (4) Innovation: new,

process, changes, and solution. (5) Citizenship:

process, human resources, law, and rules. (6) Leader-

ship: manager, boss, supervisor, and management. (7)

Performance: money, finance, sales, and salary.

The search for words and phrases that might have

a meaning in the corporate reviews is called semantic

search. Semantic search refers to understanding data

and finding matches with components of data (Bast &

Bocchuld 2016). Semantic search is frequently used

in Natural Language Processing to understand the

meaning behind large amounts of text (Bast, 2016).

Concerning search quality, there are two main

aspects: the matching between a keyword query and

a document, and the ranking of the matching

documents (Bast, 2016). With help from TSquery and

TSvector in PostgreSQL, both matching and ranking

search terms was provided in the prototype. These

data types in PostgreSQL are used for full-text search

(Patterson, 2019). The word associations were

explicitly queried in order to make it visible what

employees discussed on in reviews.

Words and phrases that were associated with

Fombrun’s dimensions were searched upon in Dutch.

The search for familiarity was a preliminary step

in performing sentiment to explore which dimensions

were involved.

Note that, importantly, before analysis, the dataset

was checked for missing data. The data was

structured with Python code that uses regular

expressions from the re module to assure that

sentiment analysis could be performed. The regular

expressions were needed to ascertain that the data was

accessible for analysis.

In addition, the Pandas library formed a basis to

perform many tasks in structuring the data. For instan-

ce, Pandas was loaded to store the data in a column.

After this process, SpaCy was used to perform

text analysis. SpaCy has grown in its popularity due

to its ease-of-use and support of Natural Language

Processing features (Broucke, 2018). SpaCy is an

open-source natural language processing library for

python. It is designed to help build applications that

process large volumes of text (Broucke, 2018).

To be able to run SpaCy, Python was installed,

also in a Docker container. The reason for this is to

provide a more consistent process.

SpaCy was used for the following three

functionalities: (1) Language recognition, most

reviews were Dutch, however it was seen that a few

reviews were also in French and English. (2)

Removing stop words, required to perform text

analysis because it eliminates unnecessary words. (3)

Tokenization, splitting the reviews into separate

words.

Note that the use of SpaCy for natural language

processing led to more accurate sentiment results.

The use of tokenization and lemmatization (spreading

sentences into words and using the root of a word)

resulted in more accurate sentiment analysis.

After using SpaCY to prepare the text from the

reviews, Pattern library was used to calculate

sentiment, which gave two values: Polarity and

Subjectivity. In Pattern, the sentiment object is used to

find the polarity (positivity or negativity) of a text

along with its subjectivity (Malik, 2020). The

sentiment score is given between 1 and -1, depending

on the positive or negative adjectives that are used.

Also, subjectivity is a value between 0 and 1. This is a

quantification of the amount of personal opinion, and

factual information found in the text (Malik, 2020).

It is useful to explore both polarity and

subjectivity because reviews are usually based on

personal opinion rather than factual information

(Ruiz-Mafea, 2018).

In addition, a Python module LangDetect was

used to recognize different languages in reviews. For

instance, Pattern can be utilized in different

languages. Therefore, the text needs to be recognized.

LangDetect could recognize whether the reviews

were in Dutch, English, or French. It should be noted

that more languages could be applied, however, in our

dataset these were not detected during processing.

After applying SpaCy and language detection, Pattern

was used to calculate sentiment.

After applying SpaCy and language detection,

Pattern was used to calculate sentiment.

To summarize, three protoypes were built in our

research: (1) Webcrawler prototype. (2) A prototype

to do full text search with data stored in a relational

database, PostgreSQL. (3) A protoype that performs

text analysis and sentiment analysis.

Figure 2: Averaged relative overall sentiment after full text

search on relevant corporate reputation words per

dimension per organization.

A Company’s Corporate Reputation through the Eyes of Employees Measured with Sentiment Analysis of Online Reviews

383

Figure 2 shows the measured division of positive and

negative reviews amongst different organizations.

Note that the organizations cannot be directly

compared with each other because they do not operate

in a similar sector and branch.

An overall observation from this graph is that

most organizations are negatively reviewed.

According to Muse, a dissatisfied employee is more

likely to post a negative review and express his

frustration than a satisfied employee (Wolf, 2019). A

satisfied employee is likely to tell three close friends

about a company's positive experience, whereas a

dissatisfied employee could feel the need to let more

people know about a bad experience in an

organization (Wolf, 2019).

Thus, this could imply that negative reviews that

are posted are usually intended to harm the organiza-

tion and affect corporate reputation (Wolf, 2019).

The graph illustrates that DHL is the worst-

reviewed organization in our case study. 81% of all

reviews from DHL are indicating as negative. On the

contrary, Coolblue is considered the best-reviewed

organization, even though 64% is negative.

4 CONCLUSIONS

In the literature review in this paper, different

approaches to measure corporate reputation were

discussed. Conducting surveys for measuring

reputation is seen as a popular method and has been

exercised frequently by researchers. However,

dimensions such as leadership, workplace, and

governance that are relevant to the stakeholder group

of employees can also be measured in online reviews.

This new approach has the potential to show

organizations that reviews on the public web can do

harm to their corporate reputation. Furthermore, we

discussed that online reviews’ perceived quality and

trustworthiness of an organization is affected if

reviews do not contain certain factors.

The results of our methodological processing

pipeline that we implemented suggest that sentiment

analysis in combination with semantic search is a

useful approach to measure corporate reputation from

public reviews on the web. We have found that a large

proportion of reviews was indicated to have negative

sentiment scores in the scraped dataset from the

public website indeed. A well-known explanation of

this finding is that employees are more likely to post

a review after an unpleasant experience in an

organization. While being aware of this, it is advised

that organizations use the applications in our toolbox

to recognize negative reviews from an early stage.

ACKNOWLEDGEMENTS

This paper has been inspired on the MSc master

project of Rim Lam-Lion who was involved via the

master Digital Driven Business at HvA. Thanks go to

Frederik Situmeang for providing some useful

suggestions to an initial version of this manuscript.

Rob Loke is assistant professor data science at

CMIHvA.

REFERENCES

Anderson. (2016, november 4). Working with Docker

containers. Retrieved from Digital Ocean:

https://www.digitalocean.com/community/tutorials/wo

rking-with-docker

Arruda, W. (2019, september 22). Why Online Employer

Reviews Are Crucial To Your Branding. Retrieved

from Forbes: https://www.forbes.com/sites/william

arruda/2019/09/22/why-online-employer-reviews-are-

crucial-to-your-branding/#1b61b9ea1f45

Bast, B. (2016). Semantic Search on Text and Knowledge

Bases. Foundations and Trends in information retrieval,

156.

BBC. (2020). Databases and Data Capture. Retrieved from

BBC: https://www.bbc.co.uk/bitesize/guides/z8yg87h/

revision/4#:~:text=Why%20use%20a%20database%3

F,edit%20or%20delete%20old%20data

Bhandari, P. (2020, 07 30). methdology. Retrieved from

Scribbr: https://www.scribbr.com/methodology/

qualitative-research/

Bronn. (2015). A Systems Approach to Understanding how

Reputation Contributes to Competitive Advantage.

Corporate Reputation review, 19.

Broucke, B. (2018). Practical Web Scraping for Data

Science. Leuven.

Choudhury. (2020, april 08). Scrapy VS Beautiful Soup: A

Comparison Of Web Crawling Tools. Retrieved from

analyticsindiamag: https://analyticsindiamag.com/scra

py-vs-beautiful-soup-a-comparison-of-web-crawling-

0tools/#:~:text=Scrapy%20is%20an%20open%2Dsour

ce%20framework%2C%20whereas%20Beautiful%20

Soup%20is,the%20developer%20what%20they%20ne

ed.

Courtright, S. (2009). leveraging organizational innovation

for stategic reputation management. corporate

reputation review, 26.

Davis. (2005). New Directions in Corporate Governance.

annual review of sociology, 20.

Fombrun. (1996). Reputation: Realizing Value from the

Corporate Image. Harvard Business Review.

Fombrun. (2015). Corporate Reputation Review:

Stakeholder tracking and analysis: The RepTrak system

for measuring corporate reputation. Macmillian

Publishers, 22.

Garbade, D. M. (2018, October 15). A Simple Introduction

to Natural Language Processing. Retrieved from

DATA 2021 - 10th International Conference on Data Science, Technology and Applications

384

Becoming Human: https://becominghuman.ai/a-

simple-introduction-to-natural-language-processing-

ea66a1747b32

Golafshani. (2003). Understanding reliability and validity

in qualitative research. the qualitative report, p. 14.

Indeed. (2020, March 30). Terms & Conditions. Retrieved

from Indeed: https://www.indeed.com/legal?hl=en&

redirect=true#api

Kircova, E. (2018). The Effect of Corporate Reputation On

Consumer Behaviour and Purchase Intention.

Management Research and Practice, 13.

Kotter. (1995). Leading Change: Why Transformatoin

efforts fail. Harvard business review, 12.

Lange, L. D. (2011). Journal of Management:

Organizational reputation. Sage, 34.

Li, S. (2019, Maart 18). A Complete Exploratory Data

Analysis and Visualization for Text Data. Retrieved

from Towards Data Science: https://towardsdata

science.com/a-complete-exploratory-data-analysis-

and-visualization-for-text-data-29fb1b96fb6a

Malik, U. (2020). Python for NLP: Introduction to Pattern

Library. Retrieved from Stack Abuse:

https://stackabuse.com/python-for-nlp-introduction-to-

the-pattern-library/

Mariconda, L. (2014). Being Known: A Literature Review

on Media Visibility, Public Prominence and Familiarity

with Implications for Reputation Research and

Management. Corporate Reputation Review, 19.

Murphy, R. (2019). Local Consumer Review Survey.

Retrieved from Brightlocal: https://www.bright

local.com/research/local-consumer-review-

survey/?SSAID=314743&SSCID=61k4_d7wyr

Nolan, G. G. (2013). Examining How Corporations Use

Online Job Ads to Communicate Employer Brand

Image Information. Corporate Reputation Review, 14.

Orlitzky, S. (2012). Assessing stakeholder satisfaction:

Toward a supplemental measure of corporate social

performance as reputation. corporate reputation review,

19.

Patterson, J. (2019, August 1). full Text searching with

Postgres. Retrieved from Forestry: https://forestry.io/

blog/full-text-searching-with-postgres/

Perino, M. (2019, juni 25). 10 American companies with

the worst reputations. Retrieved from Business Insider:

https://www.businessinsider.nl/american-companies-

with-the-worst-reputations-2019-6/

Rizvi. (2017, july 26). Web Scraping in Python using

Scrapy (with multiple examples). Retrieved from

Analytics Vidhya: https://www.analyticsvidhya.com/

blog/2017/07/web-scraping-in-python-using-scrapy/

#:~:text=Scrapy%20is%20a%20Python%20framewor

k,your%20preferred%20structure%20and%20format.

&text=Scrapy%20is%20that%20framework.,need%20

to%20reinvent%20the%20wheel.

Ruiz-Mafea, C. C.-P. (2018). The role of emotions and

conﬂicting online reviews on consumers' purchase

intentions. Journal of Business Research, 9.

Sanders, A. (2016, november 14). What Is Semantic Search

and What Should You Do About It? Retrieved from

Moz: https://moz.com/blog/what-is-semantic-search

Shayaa, J. b.-G. (2018). Sentiment Analysis of Big Data:

Methods, Applications, and Open Challenges. Digital

Object Identifier, 22.

Sheedy, G. (2018). A view from the inside. Corporate

Governance, an international review, 20.

Smith, S. W. (2010). Does brand management of corporate

reputation translate into higher market value? Journal

of Strategic Marketing, 22.

Smitha, S. W. (2010). Does brand management of corporate

reputation translate into higher market value? Journal

of Strategic Marketing, 22.

Wen, Y. R. (2014). Sentiment Analysis in MOOC

Discussion Forums: What does it tell us? Educational

data mining, 8.

Wendy Salomon. (2014, October 23). Uncommon Sense:

How Do You Measure Reputation? Retrieved from

Nielsen: https://www.nielsen.com/us/en/insights/

article/2014/uncommon-sense-how-do-you-measure-

reputation/#:~:text=Measuring%20corporate%20reput

ation%20effectively%20shows,%E2%80%9Cprogram

%2Ddriven%E2%80%9D%20discipline.

Wolf, K. D. (2019). Should You Believe Company

Reviews? Retrieved from Muse: https://www.the muse.

com/advice/should-you-believe-company-reviews

Xianghua. (2012). Multi-aspect sentiment analysis for

Chinese online social reviews based. Knowledge-Based

Systems, 10.

A Company’s Corporate Reputation through the Eyes of Employees Measured with Sentiment Analysis of Online Reviews

385