You are on page 1of 18

Available online at www.sciencedirect.

com

ScienceDirect
Journal of Interactive Marketing 53 (2021) 111 – 128
www.elsevier.com/locate/intmar

Past, Present, and Future of Electronic Word of Mouth (EWOM)


Sanjeev Verma a,⁎& Neha Yadav a
a
National Institute of Industrial Engineering (NITIE), Vihar Lake Road, Powai, Mumbai-400087

Abstract

Communication platforms are undergoing a transition from physical to digital spaces. The Internet has radically changed the business scenario
wherein people have become the “media” for collaborating to share information. Practitioners have started using electronic word of mouth
(eWOM) for consumer insight through text analytics, sentiment, hashtag analytics, and other machine learning tools. Due to the growing
importance of the topic as a research area, it is vital to map the research domain. This paper uses performance analysis tools (quantitative and
qualitative) and science mapping tools (conceptual and intellectual structures) for literature review and the identification of future research
directions. The content analysis of emergent topics and themes offers directions for future researchers. Recommender systems, data mining
(artificial intelligence and machine learning), semantics, and sentiment analysis are emergent topics. We contribute to the eWOM literature by
canvassing state-of-the-art research and proposing future research directions.
© 2020 Direct Marketing Educational Foundation, Inc. dba Marketing EDGE. All rights reserved.

Keywords: Electronic word of mouth (eWOM); Performance analysis; Conceptual structure; Intellectual structure

Introduction communications between marketers and consumers (Weisfeld-


Spolter, Suri, and Gould, 2014).
More than two decades ago, it was predicted that businesses Due to its continually increasing importance, there has been
would not compete only in the marketplace but also in the a recent spike in the number of articles published on eWOM
market space and that there would be a paradigm shift in communication (Baur and Nystrom, 2017; Hao, Ye, Li, &
business models to focus on both of these markets simulta- Cheng 2010; Lee & Youn 2009; Lee, Rodgers, & Kim 2009;
neously (Rayport & Sviokla 1994). Currently, consumers are Purnawirawan, Eisend, De Pelsmacker, & Dens 2015; Zarrad &
using multiple channels (offline and online) for becoming Debabi 2015); however, the scope of published literature is
familiar with and purchasing products. In the last decade, the large and fragmented. To consolidate learning on eWOM
Internet has radically changed the business scenario by literature, previous researchers attempted a review of eWOM
becoming an important platform wherein people become the research from different perspectives. A previous review by
“media” for collaborating to share information (Li & Wang Tuma and Decker (2013) systematically reviewed 53 studies
2011). According to Gupta and Kim (2004), websites are acting published between 1997 and 2007 and used a paradigm funnel
as “coffee shops” where people with similar interests can “find approach. Cheung and Thadani (2012) conducted a literature
and then electronically ‘talk’ to each other.” Social media review analysis on 83 articles published between 2001 and
platforms have given consumers a global platform to share 2010 and addressed methodological and theoretical underpin-
information via recommendations and opinions with the click nings. Chan and Ngai (2011) used the input-process-output
of a mouse (Xiang & Gretzel 2010). Web 2.0 has enabled one- (IPO) framework for the conceptualization of eWOM with the
to-one, one-to-many, many-to-one, and many-to-many types of help of a systematic literature review. Mishra and Satish (2016)
summarized the existing literature in the eWOM domain and
identified a few research areas with the help of theoretical
⁎ Corresponding author.
underpinnings. Most of the previous reviews used the
E-mail address: sanjeev@nitie.ac.in (S. Verma).

https://doi.org/10.1016/j.intmar.2020.07.001
1094-9968© 2020 Direct Marketing Educational Foundation, Inc. dba Marketing EDGE. All rights reserved.
112 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

descriptive or theme-based approaches, but the comprehensive object with noncommercial intention. Westbrook (1987)
performance analysis of scientific actors and science mapping defined WOM as “all informal communications directed at
of the eWOM research field is missing. other consumers about the ownership, usage, or characteristics
To fill this research gap, the present study aims to offer a of particular goods and services or their sellers.” More recently,
comprehensive performance analysis (quantitative and qualita- Barreto (2014) defined WOM as an “oral or written commu-
tive) and science mapping (conceptual and intellectual struc- nication process, between a sender and an individual or group
tures) of the eWOM research field. Performance analysis of of receivers, regardless of whether they share the same social
scientific actors such as journals, authors, documents, and network, to share and acquire information, on an informal
keywords will help in understanding the growth and impact of basis.” All three definitions discussed have treated word of
the research field. Science mapping with cooccurrence analysis, mouth as an informal communication between subjects (e.g.,
strategic maps, and thematic evolution (alluvial diagram) will sender and receiver) concerning an object (e.g., products,
help in understanding the conceptual structure of the eWOM brands, organizations, or sellers) or an experience (such as
research field. Cocitation analysis will help to identify the ownership or usage) for sharing and acquiring information with
subthemes of the eWOM research domain. Additional content noncommercial intent.
analyses of emergent topics and themes will offer research With the emergence of Web 2.0 and new media channels,
directions for future researchers. WOM has transformed into eWOM. Customers' remarks
Most of the previous studies of eWOM were authored prior concerning a product or company on the Internet are called
to 2012 and have included less than 100 articles, while this eWOM communication (Hennig-Thurau, Gwinner, Walsh, &
study included 1896 articles published between 2000 and 2020. Gremler 2004). Some researchers also started calling eWOM
This study is the most comprehensive both in terms of articles word-of-mouse (Xia & Bechwati 2010) due to the continually
included and timeframe. Moreover, with a fast-changing growing number of Internet users and the advent of Web 2.0
business environment and technological advancements, it is (Vázquez-Casielles, Suárez-Álvarez, & del Río-Lanza 2013).
important to track emerging changes for guiding the future. The Web 2.0 enabled consumers to obtain information through
findings of this study will act as a consolidated database of friends and friends of friends via social networks (Barreto
existing research of eWOM and will act as a guide for future 2015). This type of online information seeking and recommen-
research. This study makes a significant contribution to the dation has been defined in the literature as online word of
development of a comprehensive database for eWOM research, mouth (Duan, Gu, & Whinston 2008).
identification of research themes, and emergent topics for Due to the significance of electronic word of mouth in the
guiding academicians and practitioners. changing marketing environment, it is a fascinating construct
This paper is organized into six sections. Section two to study. Early research suggests that WOM is a significant
presents the background of electronic word of mouth (eWOM), influencer along with mass communication such as advertising
section three focuses on research methodology, section four (Katz & Lazarsfeld 1955). Previous studies suggest the
focuses on results, section five presents a discussion of significant role of eWOM in influencing the various stages
emergent topics and themes, and finally, section six presents of the consumer decision-making process (Davis & Khazanchi
the conclusion of the paper. 2008; Day 1971) and purchase intentions (Charlett, Garland,
& Marr 1995; Park, Lee, & Han 2007). Electronic word of
Background mouth has changed the buying environment, and customers
can now access the comparative evaluation of product
One of the primary objectives of marketing communication attributes with a single click of a mouse (Varadarajan &
is to persuade or convince customers to make a purchase. Yadav 2002).
Marketing communications enable companies to connect with
customers' imaginations for strong brand image and brand Methodology
recall to drive sales. Studies have repeatedly demonstrated a
clear association between marketing communications and a A literature review is a systematic and scientific approach
sales response. In addition to marketer-driven communication to review available research studies to understand the current
such as advertising, nonmarketer-driven communication such state of knowledge and addresses a specific research question
as word of mouth (WOM) is an alternate impactful channel of (Fahimnia, Sarkis, & Davarzani 2015). According to
communication (Keller 2007). Tranfield, Denyer, and Smart (2003), a literature review
Over the years, researchers have proposed different defini- systematically maps the existing research domain and assesses
tions of word of mouth (WOM). According to Arndt (1967), a future research direction that can further strengthen the
“WOM can be any oral and personal communication, positive knowledge. In this paper, we followed guidelines from
or negative, about a brand, product, service, or organization, in Saunders et al. (2009) in the conceptualization of a literature
which the receiver of the message perceives the sender to have review as an adaptive cycle that involves the process of
a noncommercial intention.” In the Arndt definition, word of defining relevant keywords, conducting a literature search, and
mouth involves an object such as a brand, product, service, or finally, performing analysis. We also referred to Rowley and
organization, and the subject uses a personal communication Slack's (2004) five-stage literature review approach for
channel to communicate positively or negatively about the conducting the review. Fig. 1 shows the review methodology
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 113

Fig. 1. Review methodology flowchart.

used in the present study. Finally, for conceptual and Selection of Bibliometric Databases
intellectual network structure analysis, we referred to the
approach by Fahimnia et al. (2015) for bibliometric and The two most reputed bibliometric databases are Scopus and
network (citation and cocitation) analysis. Comprehensive Web of Science (WoS). Initially, we explored both Scopus and
review protocols guided the review, and accordingly, recent Web of Science (WoS) databases to search the relevant
areas of research and future research directions were literature. Scopus had a broader coverage than WoS (Yong-
identified. Hak 2013), so we restricted our final search to the Scopus
114 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

database. Scopus is the largest abstract-based bibliometric analysis. The article screening was performed by two
database and includes more than 20,000 peer-reviewed journals researchers to ensure the suitability for inclusion in the sample.
from different publishers (Fahimnia et al. 2015). In terms of the
number of indexed journals, the Web of Science (WoS) Data Analysis Plan
databases have only 12,000 titles, while Scopus has more
than 20,000 titles. Bibliometrics is used for the quantitative and qualitative
research assessment of the academic field (Callon, Courtial, &
Laville 1991; Coulter, Monarch, & Konda 1998; Henderson,
Defining keywords (Search Strategy)
Shurville, & Fernstrom 2009; Ramos-Rodrguez & Ruz-
Navarro, 2004; van Raan, 2005). Specifically, bibliometrics is
An extensive literature search was conducted using key-
a set of methods used to study or measure texts and
words representing electronic word of mouth (eWOM). The
information, especially in large datasets. In bibliometrics,
primary and synonym keywords used for searching the
there are the two main procedures of performance analysis
literature are the following:
and science mapping (van Raan, 2005). Performance analysis
“e-WOM, Electronic Word of Mouth, Online Word of Mouth,
aims at evaluating groups of scientific actors (documents,
Internet Word of Mouth, Online Customer Review, Online
authors, journals, and countries) and their impact on the
Recommendation, Internet Recommendation, Electronic Referral,
research domain (van Raan, 2005). Science mapping aims at
Online Rating, Consumer Generated Content, Online Opinion,
displaying the structural and dynamic aspects of scientific
User Generated Content, User ratings, User Comments, and User
research. A science map is used to represent the cognitive
Review”.
structure of a research field. Science mapping can be done
Boolean operators “OR” were used in the “title, abstract,
using cocitation analysis (Small 1973) and coword analysis
keywords” tab of the Scopus database for searching the research
(Callon, Courtial, Turner, & Bauin 1983).
papers. Initially, a total of 12,367 articles were retrieved from the
We have used a combination of performance analysis and
search result.
science mapping for detecting and visualizing the conceptual
subdomains of the eWOM research field. This allows us to
Refining the Initial Results (Inclusion and Exclusion Criteria) quantify and visualize the thematic evolution of the research
The search results were delimited with the help of inclusion field. The performance measure is divided into two categories
and exclusion criteria for the extraction of the most relevant of quantitative and qualitative measures. For quantitative
articles for the literature review. Given the research objective, measures, the productivity of the scientific actors (documents,
the search results were restricted to articles, articles in the press, authors, journals, and countries) are analyzed, whereas
and reviews. Conference papers, book chapters, commentaries, qualitative measures show the quality-based indices. Frequency
erratum, and so on were excluded from the search results since computations for various fields of the study indicated the
articles and reviews published in journals are considered to be significance of authors, documents, sources, affiliations, and
“certified knowledge” (Ramos-Rodríguez and Ruiz-Navarro, countries (Garfield 1972). Qualitative measures total global
2004). After the application of the first delimiter, the number of citations (TGC) and total local citations (TLC)) measure the
articles was reduced to 6,041. number of citations received and bibliometric indices such as
The search results were then delimited on the second the h-index (Alonso, Cabrerizo, Herrera-Viedma, & Herrera
exclusion criteria of language. Only articles published in 2009; Cabrerizo, Alonso, Herrera-Viedma, & Herrera 2010;
English were included in the final set of articles for further Hirsch 2005). TGC indicates the overall impact of the scientific
review. After the application of the second delimiter of the factor on the specific research field, while TLC denotes the
English language, the number of articles was reduced to 5,741. number of citations within the same literature set. In
The third exclusion was done based on the publication conjunction with TGC and TLC, total local citation per year
timeframe. Articles were delimited for publication between (TLC/t), global citation score per year (TGC/t), and trending
2000 and 2020 because information technology accelerated in patterns of total local citations (TLCe) are also reported in this
the 1990s, and, initially, information technology was used for study. TGC and TLC indicate the significance and impact of
online searches. Online reviews and eWOM accelerated as a the scientific actors (documents, authors, journals, and
discipline at the beginning of the new millennium; thus, search countries) in the research field (Culnan 1986; Sharplin &
results were delimited on a timeframe (i.e., between 2000 and Mabry 1985). Although there are contrary views on using
2020). After the application of the third delimiter of the citation analysis as a benchmark for the research impact of an
timeframe, the number of articles was reduced to 5,709. The article, citation analysis is the widely used technique for
fourth and final delimitation was done based on excluding literature analysis and identification of significant research
articles from nonrelated fields; articles were limited to business, works in the research field (Mac Roberts & Mac Roberts 1989;
management, and accounting. The final set of search results Vokurka 1996).
included 1896 articles, and these search results were exported Science mapping or bibliometric mapping is a spatial
to.csv and Bibtex format files for future analysis. A detailed representation of how disciplines, fields, specialties, and
examination of 1896 articles was performed by reading the individual papers or authors are related to one another (Small
abstract and full paper to assess the suitability of the articles for 1999). It is focused on monitoring a scientific field and
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 115

delimiting research areas to determine its cognitive structure Results


and evolution (Noyons, Moed, & van Raan 1999). Cocitation
and coword analysis are commonly used to map science (Small Due to the emergence of information technology and various
2006). Cocitation analysis maps the structure of a research field meta-search engines such as Google Scholar and Microsoft
through pairs of documents that are commonly cited together Academic Search, researchers are better equipped to obtain a
(Coulter et al. 1998), while coword analysis is a content larger picture of a research domain. However, most of the freely
analysis technique to map the strength of association between available bibliometric tools have their limitations. For instance,
information items in textual data (Callon et al. 1983). Coword Google Scholar provides extensive coverage of the research
analysis directly addresses sets of terms shared by documents, works published in the research domain, but Google's search
mapping the pertinent literature directly from the interactions of result may include conference papers, magazine articles, book
key terms. Thus, coword analysis allows us to discover the chapters, white papers, and trade magazine publications along
main concepts treated by the field; it is a powerful technique for with journal articles. For the literature review, the delimitation of
discovering and describing the interactions between different only journal articles is a tedious task in the Google Scholar or
fields in scientific research. Although both techniques are Microsoft Academic Search platforms. A significant amount of
useful for mapping science, our approach aims to discover the manual intervention is required to purify the search results. Thus,
conceptual structure and evolution of a research field; therefore, we used the most scientific and established bibliometric database,
coword analysis is more suitable. SCOPUS, to search the relevant literature in this study.
The methodological foundation of coword analysis is the Moreover, bibliometric analysis software tools such as
idea that the cooccurrence of key terms describes the content Publish or Perish, HistCite, BibExcel, and VOSviewer also
of the documents in a file (Callon et al. 1991). According to had some limitations. For instance, Gephi requires Java 1.8 or
Krsul (1998, p. 80), “this technique illustrates associations higher for running its analysis. HistCite works only with Web
between keywords by constructing multiple networks that of Science (WoS) data and Publish or Perish works only with
highlight associations between keywords, and where associa- Google Scholar and Microsoft Academic Search data. Simi-
tions between networks are possible”. In this paper, these larly, there are limitations on input and output file formats with
networks are associated with themes. Each publication in the these bibliometric tools. Computational limitations and a
field can, in turn, be characterized by a sublist of key terms, maximum number of fields in the input file add to the
which are similar to DNA fingerprints of these published complexity. We have used the R language in this study to
articles (Börner, Chen, & Boyack 2003). According to Börner overcome all these limitations. The R language is highly
et al. (2003), “by matching keyword-based fingerprints, one flexible in handling data from different databases such as
can measure the similarity between a pair of publications. The Scopus and offers extensive data analysis. R program codes can
more keywords two documents have in common, the more be modified easily to suit the research purpose, and output files
similar the two publications are, and the more likely that they can be further used with network analysis tools such as
come from the same research field or research specialty at a VOSviewer and Pajek.
higher level. Following the DNA metaphor, if two publica- The descriptive statistics of the data (Table 1) show that this
tions' ‘fingerprints’ are similar enough, they are bound to study included 1896 documents published in 397 sources. We
come from the same species.” With a list of the important have included documents published only in journals. Of the
keywords of the research field, a graph can be built where the total 1896 articles, 1777 were articles and the remaining 119
keywords are the nodes and the edges between them represent were review articles. Articles under review were published
their relationships. Two nodes (keywords) are connected if
they are presented in the same documents. We can add a Table 1
weight to each edge to represent the importance of the Descriptive Statistics of the Data.
associated relationship in the whole corpus. As a result of the Description Results
coword analysis, a set of detected themes is obtained for each Documents 1896
subperiod studied. Sources (journals, books, etc.) 397
To represent the results visually, we have used strategic Keyword plus (ID) 2,494
diagrams and an alluvial diagram. Strategic diagrams (Callon Author's keywords (DE) 4,756
Period 2000–2020
et al. 1991) helped to show the research field and its detected
Average citations per document 37.06
subfields while an alluvial diagram (Rosvall & Bergstrom Authors 3,815
2010) presented the evolution of detected clusters in Author appearances 5,154
successive periods. Coword analysis in a longitudinal Authors of single-authored documents 206
framework (Garfield 1994) presented through an alluvial Authors of multiauthored documents 3,609
Single-authored documents 247
diagram tracked the evolution of the eWOM research field
Coauthors per document 2.72
during consecutive periods. There are different degrees of Collaboration index 2.19
continuance that can be measured with various similarity
measures. We used the stability index (Braam, Moed, & van Document types
Raan 1991; Small 1977) to measure the continuance among Articles 1777
Reviews 119
clusters.
116 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

Arcles Expon. (Arcles) Table 2


450
Most Productive Journals.
400 409

350 Sources NP TC TC/t


300 Decision Support Systems 73 4,567 415.2
250 262 Tourism Management 54 4,873 443.0
200 193191 International Journal of Hospitality Management 50 3,061 306.1
150 151 Journal of Business Research 43 2,437 187.5
119128 Electronic Commerce Research and Applications 38 1,507 137.0
100 98
73
50 50
32 37
10 5 5 6 17
0 4 4
1995 2000 2005 2010 2015 2020 (number of published papers, NP) shows Decision Support
System has published the greatest number of articles (73). The
Fig. 2. Number of Articles Published per Year. higher numbers of eWOM publications in Tourism Management
(54) and the International Journal of Hospitality Management
between 2000 and 2020. Most of the documents have multiple (50) show the widespread application of eWOM in the tourism
authors, and only 247 documents have a single author. On and hospitality sector. The Journal of Business Research and the
average, every document has been coauthored by 2.72 authors Electronic Commerce Research and Applications journal also
with a collaboration index of 2.19. The average citation per appeared in the top five sources. This indicates the growing
document (37.06) is high, which indicates the popularity of the interest of coupling between eWOM, business, and ecommerce
research field. Keyword plus and author keywords are two performance.
scientific units for cooccurrence analysis. A comparative study of The qualitative measures such as total citations (TC) and
keyword plus and author plus proved keyword plus is as effective total citations received per year (TC/t) provided evidence of the
as author keywords in bibliometric analysis for mapping the growing conceptual and intellectual structure of the research
conceptual structure of a research field (Zhang et al. 2016). The field. Tourism Management received the greatest number of
keyword plus sample for our data is 2,494 keywords. citations (4873), followed by Decision Support Systems.
To understand the development of the eWOM research field, a Tourism Management is one of the best sources to publish
longitudinal analysis was performed. Fig. 2 presents the growth eWOM research as it has the highest TC/t (443). Decision
trajectory of the discipline and indicates its growing popularity. Support Systems, Tourism Management, and International
There is an exponential growth in the documents published Journal of Hospitality Management are the most preferred
between 2000 and 2019. During the last five years, the research journals for eWOM research as they excelled in all three
field's growth has been phenomenal, growing at a rate of three parameters, i.e., the number of papers, total citations, and
hundred percent. By dividing the timeframe in four equal quality of the publication.
intervals, we found the discipline develop slowly in the
beginning (between 2000 and 2005) and its growth accelerated Author Performance
in the second subperiod (between 2006 and 2010). Only 28
documents were authored between 2000 and 2005 while the The h-index, total citations (TC), and number of articles
number was five times that (142 documents) during the following authored (NP) computations indicated the most influential
five years (between 2006 and 2010). The growth trajectory authors (Table 3). Law authored the highest number of articles
accelerated in the last decade, with 569 documents published (26) followed by Xu (11), Duan (10), Gu (10), Bilgihan (8), and
between 2011 and 2015 and 1,055 documents published during Schuckert (8). The h-index score also indicates the significant
the last four years (between 2016 and 2019). The growth pattern contributions of Law (60), Bilgihan (24), and Gu (21). The h-
suggests the relevance and implication of the research field are index and number of publications indicate the involvement of
increasing continuously. The contemporary trend and evolution the author in a research domain, but it does not necessarily
of the discipline can unfold new paradigms for future researchers.
Table 3
Performance Analysis Most Influential Authors.
Author Author Affiliation NP H-index TC
The performance analysis of different scientific actors offers
the structure and development of the eWOM research field. The Law, Rob The Hong Kong Polytechnic 26 60 15,631
University, Kowloon, Hong Kong
various scientific actors used for the analysis are journals, Xu, Xun California State University 11 10 449
articles, authors, and keywords. Both quantitative and qualita- Stanislaus, Turlock, United States
tive aspects of performance analysis provided vital insights for Duan, Wenjing The George Washington University, 10 14 3,010
future researchers. Washington, D.C., United States
Gu, Binx W. P. Carey School of Business, 10 21 4,121
Tempe, United States
Source Performance Bilgihan, Anil Florida Atlantic University, Boca 8 24 2082
Raton, United States
Table 2 presents the top five sources that included research Schuckert, The Hong Kong Polytechnic 8 16 926
Markus R. University, Kowloon, Hong Kong
publications of eWOM research. The quantitative measure
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 117

translate to the popularity of an author due to the quality of the looking into keywords used by research papers and their
publication. TLC/t is an indicator of the quality of the publication aggregation (Scott & Lane 2000; Sidorova, Evangelopoulos,
and indicates the popularity gain in minimum time. Gu scored the Valacich, & Ramakrishnan 2008). The words most frequently
highest on the TLC/t parameter (206.7), which indicates that every used by the previous author were computed and presented in
article authored by B. Gu received the greatest number of citations Fig. 3. For the identification of top keywords, both keywords
per year among the top twenty authors. used by previous researchers and indexed keywords were
considered. The intersection of author keywords and indexed
Documents Performance keywords were included for analysis.
As we can see from Fig. 3, previous researchers have used
As discussed above, qualitative measures such as total keywords such as including user-generated content, electronic
global citations (TGC) and total local citations (TLC) helped in word of mouth, and online reviews for capturing the eWOM
the identification of the most influential research papers. TLC phenomenon. User-generated content is the most frequently
represents the research papers that are cited the greatest number used keyword (108 times) followed by electronic word of
of times by the group of documents analyzed in this study, and mouth (46 times) and online review (32 times). The most
TGC indicates the documents most cited globally, i.e., that frequently used keywords also include the channels used for
received citations from any document within or outside the sharing eWOM, data mining techniques, and expected out-
collection of papers in this study. Table 4 presents the top five comes of eWOM. eWOM spreads through online channels and
locally and globally cited documents based on TLC, TLC/t, social networking sites such as Facebook and Twitter. Social
TGC, and TGC/t computations. networking primarily focuses on exchanging information
Litvin et al. (2008), Kaplan and Haenlein (2010), Forman et al. between members of networks, but social media is an online
(2008), Vermeulen and Seegers (2009), and Sparks and media to communicate to a larger number of people.
Browning (2011) are the top five articles according to the TLC eWOM is a rich source of consumer insight, and marketers are
ranking. Kaplan and Haenlein (2010) is the key research paper mining it for their advantage. Keyword analysis of existing
that drives the eWOM research domain and received the most literature suggests the usage of data mining to aid decision
global citations (5568). In this paper, Kaplan and Haenlein making and effective marketing strategies. Data mining is one of
(2010) propagated the use of social media for harnessing business the most frequently used words, and emerging technological
excellence and provided a classification of social media based on advancements such as artificial intelligence have improved
applications such as collaborative research projects, blogs, marketing intelligence for deeper consumer insights. Most
content communities, social networking sites, virtual game frequently used words also include outcome-driven keywords
world, and virtual social world. Litvin et al. (2008) is another such as marketing (60 times), decision making (44 times),
very popular research paper based on the total local and total consumer behavior (64 times), and sales (84 times). In the
global citations. Litvin et al. (2008) propounded electronic word outcome related constructs, the sale is the predominant keyword
of mouth as a new cost-effective method of capturing, analyzing, (33%), which shows the significance of eWOM in driving sales.
interpreting, and managing influence in the digital era that should
be treated cautiously. Cantallops and Salvi (2014) is another Thematic Analysis and Strategic Diagram
noticeable research paper that became prevalent in a short period We performed coword analysis on indexed keywords to map
(TLC/t = 21.20). In this paper, Cantallops and Salvi (2014) the conceptual structure of eWOM research. The themes are
reviewed research on eWOM in the hotel industry published presented on a strategic diagram in a two-dimensional space as per
between 2008 and 2013 and identified two research streams— their centrality and density. Centrality measures along the x-axis
antecedents of eWOM and impacts of eWOM from the consumer and density measures along the y-axis. A rank value based on the
and organization perspectives. median is preferred due to its legibility (Cahlik 2000) for centrality
and density computation. Different clustering algorithms are used
Keywords to create the strategic thematic map in extant literature (Callon et

Performance Occurrences
In this section, we present the results of our keyword 140 127
120 108 104
analysis. This type of discussion assists in revealing the 100 86 84
intellectual core and identity construction of the discipline by 80 63 60 64
60 46 43 44
32 38
40
Table 4 20
Most Influential Research Papers. 0
s
th

Document TLC TLC/t TGC TGC/t


ou
m
of

Litvin, Goldsmith, and Pan (2008) 226 20.55 1,084 98.55


Kaplan and Haenlein (2010) 142 15.78 5,568 618.67
Forman, Ghose, and Wiesenfeld (2008) 132 12.00 791 71.91
Vermeulen and Seegers (2009) 132 13.20 561 56.10
Sparks and Browning (2011) 131 16.38 555 69.38
Fig. 3. Most Frequent Words.
118 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

al. 1991). Coulter et al. (1998) proposed a simple center algorithm eWOM from an information management perspective. Past
to compute coword analysis and to prepare a strategic diagram. studies explored and empirically tested eWOM as an informa-
Due to its widespread prevalence, popularity, and advantages, we tion source in a narrow sense and failed to extrapolate it for
have used a simple cluster algorithm for strategic mapping in this wider generalizability. A limited view of eWOM as a source for
study. information dissemination and data acquisition delimited the
Based on the position of themes on the strategic map along applicability of the research field.
its two-axes (x-axis denotes centrality and y-axis denotes Themes in the lower-left quadrant are weakly developed and
density), themes can be classified into four quadrants (Cahlik marginal. The themes of this quadrant have low density and
2000; Callon et al. 1991; Coulter et al. 1998). Themes in the low centrality, mainly representing either emerging or
upper-right quadrant are both well developed and important for disappearing themes. Decision-making and social media
the structuring of a research field. They are known as the themes are placed in this quadrant of the strategic map. Initial
motor-themes of the specialty, given that they present strong studies of eWOM focused on channels such as the Internet and
centrality and high density. The placement of themes in this social media for exchanging information. The larger social
quadrant implies that they are related externally to concepts media theme sphere points at wider coverage of social media
applicable to other themes that are conceptually closely related. studies for exchanging information.
As depicted in Fig. 6, recommender system and data mining are Themes in the lower-right quadrant are important for a research
in the high centrality and high-density quadrant. This quadrant field but are not developed. Thus, this quadrant presents
represents closely related keywords used in the most popular transversal and basic themes. User-generated content and
and widely published research studies. electronic commerce are present in the lower-right quadrant
Recommender systems and data mining are important (Fig. 4). User-generated content signifies the exchange of
subfields of eWOM research. Data mining is relatively higher information among users based on their experience with products
in centrality than the recommender system. This implies that and services. Since user-generated content is exchanged without
data mining is a much more popular technique for understand- filters and conveys the first-hand experience of users, it is
ing eWOM, its antecedents, and its consequences. Coword considered to be more trustworthy among consumers. User-
analysis of the data mining theme also include keywords such generated content is a popular research topic as the information
as text mining, semantic analysis, and sentiment analysis. This exchanged on social networks is the original feeling and
clearly shows the usage of both quantitative and qualitative experience of the connected users in close-knit social networks.
approaches for mining eWOM information. The data mining Past literature also points to the association between eWOM and
theme sphere is also larger than the recommender system theme ecommerce performance. Both themes (user-generated content
sphere, which also validates the heightened popularity of the and ecommerce) in the lower right quadrant contributed
data mining technique in eWOM research. significantly to the progression of eWOM research. Due to a
Themes in the upper-left quadrant have well developed lack of theoretical rigor, studies focused on UGC and ecommerce
internal ties, but unimportant external ties and so are of only could not develop fully and remain transversal.
marginal importance for the field. These themes are very
specialized and peripheral. Fig. 6 indicates theme management Thematic Evolution
and theme web are placed in the upper left quadrant. The To understand the thematic evolution of eWOM research
smaller size of thematic spheres presented in the upper left over the last two decades, the raw data were divided into four
quadrant points at a lesser number of studies attempted on time periods (subperiods). R-Script divided the total number of

Fig. 4. Thematic Analysis.


S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 119

Fig. 5. Thematic Evolution (Alluvial Diagram).

documents into four zones automatically based on the number the theme belongs to more than one thematic area. However, some
of documents. The total 1896 documents under bibliometric themes do not have a shadow, which implies that these themes do
review were divided into four time periods, i.e., 2000–2013 not belong to any thematic area. As depicted in the alluvial diagram,
(460 documents), 2014–2016 (472 documents), 2017–2018 the eWOM research field presents great cohesion. Most themes are
(453 documents), and 2019–2020 (511 documents). grouped in some thematic areas and originate from a theme in a
In Fig. 5, the alluvial diagram portrays the thematic previous subperiod. Thematic areas such as decision making and
evolution of the eWOM research domain. In different color- recommender systems evolve continuously, which implies their
shadows groups, the themes belong to the same thematic area. continually increasing importance. Some thematic areas such as the
Some themes have more than one shadow, which implies that Internet, ecommerce, and user-generated concepts follow a

Time Period 1: 2000-2013 (460 documents) Time Period 2: 2014-2016 (472 documents)

Time Period 3: 2017-2018 (453 documents) Time Period 4: 2019-2020 (511 documents)

Fig. 6. Strategic Map of Different Time Periods.


120 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

sinusoidal graph pattern. The thematic area representing the To measure the thematic overlap between two consecutive
decision support system disappears after the first subperiod. subperiods, the stability index is measured (Small 1977). The
Thematic areas such as such as electronic commerce and social stability index for the recommender system for the third
networking increasingly evolve. Themes such as sales, social subperiod (2017–2018) and fourth subperiod (2019–2020) is
networking, and data mining emerged from the ecommerce theme. 0.65, which is significant for shared keywords. The stability
Furthermore, three thematic areas (decision making, UGC, index for user-generated content is 0.58, social networking is
and ecommerce) are present in all the studied subperiods. These 0.48, and data mining is 0.42, showing that these themes have a
three thematic areas have maintained the eWOM researchers' consistent evolution in the eWOM research domain.
interest in all the studied subperiods. Looking at the
progression of the research fields, the number of themes in Cocitation Network Analysis
the final subperiod (2019–2020) has increased compared to Cocitation analysis is primarily used to understand the
previous subperiods. During the last subperiod of 2019–2020, interrelationships between existing research studies based on
eWOM research was classified into nine thematic areas authors and documents (Small 1973; Pilkington and Catheriene,
indicating context-specific studies. Marketing, electronic word 1999). The network analysis of research studies is called
of mouth, websites, and information systems themes evolve cocitation analysis and depicts the intellectual structure of the
during the last subperiod. research domain (Chen, Ibekwe-San Juan, & Hou 2010).
A strategic map for each subperiod (Fig. 6) points at the Cocitation network analysis works on the data clustering
development of various themes periodically in terms of their principle. In the data clustering technique, research papers in
centrality and density. The strategic map of the first subperiod the same cluster share similar characteristics (Radicchi,
(2000–2013) shows the Internet garnered significant attention Castellano, Cecconi, Loreto, & Parisi 2004). Accordingly,
initially, and various studies attributed the increase in eWOM to different clusters are formed to understand the subthemes of
the growth of the Internet. The Internet theme occupies the high study in the research domain. Nodes represent research papers
centrality-high density space initially. Recommender systems and edges represent the connections between research studies.
and decision making were less evolved in the beginning. A The thickness and density of the edges indicate the similarity
decision support system was a nascent concept during the index between research studies (Clauset, Newman, & Moore
2000–2013 subperiod. User-generated concepts and electronic 2004; Leydesdorff 2011; Radicchi et al. 2004). The modularity
commerce were transversal concepts. measures the strength of the relationship between nodes, and its
In the second subperiod (2014–2016), online social value varies between − 1 and + 1 (Blondel, Guillaume,
networking became the motor theme and occupied the high Lambiotte, & Lefebvre 2008). The modularity index is a widely
centrality-high density quadrant. eWOM became the center of used and accepted concept for cocitation analysis (Mishra,
attention, and various studies explored the impact of eWOM on Gunasekaran, Childe, & Papadopoulos., & Dubey, R. 2016).
organizational performance. The organizational performance The modularity index can be calculated with the help of
was gauged on both economic (sales) and noneconomic (brand various algorithms such as the edge between algorithms,
equity) performance parameters. The impact of eWOM on sales Louvain algorithm, leading eigenvalues, spin glass algorithm,
was a traversal concept during the second subperiod. The and walk trap algorithm. In this study, we used the Louvain
Internet was treated as the basic theme in the second subperiod algorithm to calculate the modularity index. The Louvain
and few specialized studies of data mining were found. algorithm is an optimization method that appears to run in time
In the third subperiod (2017–2018), social media become a O (n log n). The optimized value is the modularity, and its value
popular basic theme. eWOM, social media, and sales lies between − 1 and + 1. The modularity measures the density
relationship studies became transversal due to higher centrality of links within the community compared to links between
index values. Recommender system and user-generated content communities. For a weighted graph, the modularity is
themes occupied the right-hand quadrant of the strategic map, calculated as follows:
depicting high centrality and high density. This indicates that
user-generated content and recommender system themes  
1 ∞ kik j
become the major contributor of eWOM research during the Q¼ ∑ij aij − ðCiCjÞ
2m 2m
2017–2018 subperiod. This period also marks the presence of
specialized studies of social networking and decision-making
themes. In this study, the Louvain algorithm captured a 50-node
Recently (2019–2020), social media has become the basic network, and the cocitation network was drawn to identify
theme of studies, while eWOM and user-generated content subthemes of the research domain (Fig. 7). Betweenness
have become transversal themes. The recommender system centrality was calculated for nodes in different clusters to
theme occupies the upper right-hand quadrant space of the identify the most prominent research papers in different
strategic diagram, showing the growing popularity of the clusters. Usually, research papers in a cluster cite together,
recommender system theme. The growing spherical size and these research papers share the same area of interest
representing the recommender system shows the growing (Hjørland 2013). Nodes in each cluster were analyzed carefully
interest of researchers in this theme. with the help of betweenness centrality to identify the most
prominent research papers; five topmost nodes (research
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 121

Fig. 7. Co-Citation Analysis.

papers) represented the whole cluster. Each cluster was description of the content analysis of each subfield (theme) is
analyzed to identify the focus of the cluster. presented in the discussion section.
The cocitation network layout can take different structures such
as circle, Fruchterman and Reingold, Kamada and Kawai, and Discussion
multidimensional scaling layouts. We used an automatic layout
that uses the association strength algorithm. In the cocitation Cooccurrence analysis and a strategic map offered the
analysis map, strongly connected nodes occupy the center of the conceptual structure of the eWOM research field and cocitation
cluster while loosely connected nodes shift to the periphery of the analysis offered its intellectual structure. We performed a
cluster. The isolated nodes in the cocitation map represent outliers content analysis on both conceptual and intellectual structure
and the data clusters do not include these nodes. themes. As discussed in section 4.2, the strategic map suggests
Data clustering was performed and tabulated for clear the recommender system and data mining are the most
delineation of nodes in different clusters. As evident from Fig. promising emerging topics. Additionally, cocitation analysis
7, the cocitation network indicates three data clusters repre- indicated three clusters as important subthemes of eWOM
sented by three different colors. Red denotes cluster 1, green research. Thus, we focused our discussion on these emerging
denotes cluster 2, and blue indicates cluster 3. Table 5 presents topics and themes in this section.
the top five research papers from each cluster.
Content analysis of research papers from each cluster Cluster 1: Conceptualization of eWOM.
suggests the primary focus of each cluster. Cluster 1 focused
on the conceptualization of eWOM, while cluster 2 focused on Researchers in cluster 1 focused on the conceptualization of
the role of eWOM in influencing sales. Cluster 3 revolves electronic word of mouth. Brown et al. (2007) explored the
around sector-specific applications and the implication of prevalence of word of mouth communication in online
eWOM. In cluster 3, most of the studies focused on the communities and conceptualized an online social network.
experience dominant hospitality and tourism sectors. A detailed According to Brown et al. (2007), underlying theories for
122 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

Table 5
Cocitation Analysis.
Cluster No. Author and Year of Publication Title of Paper
1 Hennig-Thurau et al. (2004) Electronic Word-Of-Mouth via Consumer-Opinion Platforms: What Motivates Consumers to Articulate
Themselves on the Internet?
1 Dellarocas (2003) The Digitization of Word of Mouth: Promise and Challenges of Online Feedback Mechanisms
1 Bickart and Schindler (2001) Internet Forums as Influential Sources of Consumer Information
1 Cheung, Luo, Sia, and Chen (2009) Credibility of Electronic Word-of-Mouth: Informational and Normative Determinants of On-Line Consumer
Recommendations
1 Brown, Broderick, and Lee (2007) Word of Mouth Communication within Online Communities: Conceptualizing the Online Social Network

2 Chevalier and Mayzlin (2006) The Effect of Word of Mouth on Sales: Online Book Reviews
2 Zhu and Zhang (2010) Impact of Online Consumer Reviews on Sales: The Moderating Role of Product and Consumer Characteristics
2 Forman et al. (2008) Examining the Relationship between Reviews and Sales: The Role of Reviewer Identity Disclosure in
Electronic Markets
2 Liu (2006) Word of Mouth for Movies: Its Dynamics and Impact on Box Office Revenue
2 Mudambi and Schuff (2010) What Makes a Helpful Online Review? A Study of Customer Reviews on Amazon.com

3 Litvin et al. (2008) Electronic Word-of-Mouth in Hospitality and Tourism Management


3 Ye et al. (2009) The Impact of Online User Reviews on Hotel Room Sales
3 Vermeulen and Seegers (2009) Tried and Tested: The Impact of Online Hotel Reviews on Consumer Consideration
3 Sparks and Browning (2011) The Impact of Online Reviews on Hotel Booking Intentions and Perception of Trust
3 Ye, Law, Gu, and Chen (2011) The Influence of User-Generated Content on Traveler Behavior: An Empirical Investigation on the Effects of
E-Word-of-Mouth to Hotel Online

explaining the offline word of mouth are inappropriate to on sales, various contexts such as Amazon.com, online book
describe eWOM and its impact on consumer behavior. In social reviews, and box office revenue were considered. Liu (2006)
networks, individuals behave such as a communication channel studied the impact of eWOM on box office revenue. Tracking
and individual identification acts as a social proxy in online eWOM during the prerelease and opening week of a movie
social networks. Bickart and Schindler (2001) discussed the helps to understand the explanatory power of eWOM.
role of Internet forums as an influencer for consumer behavior. According to Liu (2006), the volume of eWOM reviews has a
Experiments on a real-life case study prove that user-generated higher explanatory power than the valence of the reviews and
content is more influential than marketer-sponsored communi- proved that eWOM impacts box office revenue positively.
cations (Bickart & Schindler 2001). Given the overarching Chevalier and Mayzlin (2006) studied the effect of eWOM of
influence of electronic word of mouth, Dellarocas (2003) sales of books on ecommerce platforms. Chevalier and Mayzlin
discussed the promises and challenges of online feedback (2006) postulated that consumers do not rely on the volume of
mechanisms. According to Dellarocas (2003), the bidirectional reviews; rather, consumers look for the quality of information
characteristics of online feedback fosters trust and confidence shared in an online review. The content and description in
among consumers. To leverage the benefit of electronic word of online reviews are more influential than the number of reviews.
mouth, organizations are using it for brand building, customer Mudambi and Schuff (2010) studied the influence of eWOM
acquisition, and retention (Dellarocas 2003). Hening-Tharou et and tried to investigate the helpfulness of reviews. According to
al. (2004) and Cheung et al. (2009) discussed the determinants Mudambi and Schuff (2010), review extremity, review depth,
of electronic word of mouth. Hening-Tharou et al. (2004) and product type are the factors that determine the helpfulness
studied electronic word of mouth through consumer opinion of a review. In examining the relationship between reviews and
platforms and tried to understand the underlying motivators for sales, Forman et al. (2008) tried to understand the role of
consumers' expressions on the Internet. The desire for social reviewer identification in the whole process. Their findings
interaction and concern for other consumers are overt suggest that reviewer identification improves online product
motivators for electronic word of mouth. In contrast, the desire sales (Forman et al. 2008). Zhu and Zhang (2010) explored the
for economic incentives and the elevation of self-image are moderating effect of product and consumer characteristics in
covert motivators. Cheung et al. (2009) assessed informational the dyadic relationship between eWOM and sales. According to
and normative determinants to assess the credibility of Zhu and Zhang (2010), firms' online marketing strategies
electronic word of mouth. The level of consumer involvement should incorporate product and consumer characteristics while
and prior knowledge moderates the influence of electronic devising eWOM strategies for improved sales performance.
word of mouth (Cheung et al. 2009).
Cluster 3: eWOM, Hospitality, and Tourism.
Cluster 2: eWOM and its Impact on Sales.
Research papers in cluster 3 revolve around the role of
The primary emphasis of research cluster 2 is to find the eWOM in experience-dominated sectors such as hospitality and
impact of eWOM on sales. To understand the effect of eWOM tourism. Experience-dominant services are intangible and
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 123

difficult to evaluate before consumption, so they rely heavily multicriteria based fuzzy systems can improve the accuracy of
on word of mouth. Litvin et al. (2008) studied the influence of recommendations (Palanivel & Sivakumar 2011).
eWOM as an economical means for marketing communication
in hospitality and tourism. Ye et al. (2009, 2011) used a log- Emergent Topic 2: Data Mining.
linear model to study the effect of eWOM on hotel room
bookings and found a 10 % increase in travelers' review ratings With the advent of Web 2.0, electronic word of mouth on
can improve hotel room online booking by more than 5%. social networking sites became the source for exchanging
Vermeulin and Seegers (2009) studied the impact of hotel ideas and experiences among users (Netzer, Feldman,
reviews on consumers' consideration with the help of Goldenberg, & Fresko 2012; Ravi & Ravi 2015). With the
experimentation methods. They found review valence, hotel availability a significant amount of user-generated content
familiarity, and the reviewer's expertise have a significant (big data), it has become an excellent source for mining
impact on consumer consideration. Sparks and Browning consumer sentiments and latent needs (Tirunillai & Tellis
(2011) used experimentation design to understand the effect 2014). Consumer-generated content is a new information
of review target (core or interpersonal), review valence source that encapsulates all stages of the consumer purchase
(positive or negative), review framing, and numerical rating cycle (Guo, Barnes, & Jia 2017), and it can help in
on influencing a consumer's hotel choice. understanding the drivers of consumer satisfaction/dissatis-
faction (Berezina et al. 2016). User-generated content
provides the opportunity to extract market structures and
Emergent Topic 1: Recommender system. competitive landscape insights (Netzer et al. 2012).
Research in the eWOM domain has largely focused on
Newly emerging location-based and event-based social quantitative ratings (Guo et al. 2017). For instance, numeric
network services offered a new platform to understand users' measurements representing volume and valence of eWOM can
preferences based on their activity history (Yin, Cui, Sun, Hu, model the impact of product reviews on sales. However, rich
& Chen 2014). User ratings based on past experiences form the information embedded in eWOM cannot be captured in scaler
basis for recommender systems in ecommerce (Guo et al. terms (Archak, Ghose, & Ipeirotis 2011). Advanced techniques
2014). Due to multiple information channels, consumers are such as linguistic analysis can delve deeper to draw rich
overloaded with information, which adds to the complexity of consumer insights (Guo et al. 2017). Text mining, opinion
selecting the correct products (Baum & Spann 2014). The mining, and sentiment analysis are better determiners of
magnitude of the problem increases when people travel to new consumers' choices than scaler valence and volume of reviews
and unknown places. The recommender system based on user (Archak et al. 2011; Ravi & Ravi 2015).
ratings solves the problem (Qui et al., 2016) and improves To demonstrate the use of text mining in building a
decision making through predictive performance based on prior consumer choice model, Archak et al. (2011) used text mining
ratings (Guo et al. 2014). The location content-aware to decompose textual reviews into segments. The text analytics
recommender system (LCARS) offers useful recommendations based on text mining techniques can offer drivers of customer
for venues or events based on personal interests and local satisfaction and dissatisfaction (Berezina et al. 2016). Semantic
preferences (Yin et al. 2014). Thus, the recommender system is characteristics of online user reviews improve the ability to
useful not only to travelers but also to local residents. diagnose the consumer characteristics and their latent needs
A recommender system uses different modeling algorithms (Cao, Duan, & Gan 2011). Guo et al. (2017) used the latent
to nest the user ratings to offer useful recommendations. The Dirichlet analysis (LDA) to identify the latent needs of
location content-aware recommender system (LCARS) works customers to serve them better. The text-mining approach,
on the two components of offline modeling and online combined with semantic network analysis tools, helps in
recommendation. Offline modeling (LCA-LDA) learns the understanding market structure (Netzer et al. 2012).
interest of individual users by mapping cooccurrence patterns. Data mining offered multifaceted advantages to both
At the same time, online recommendation decodes the user academics and practitioners. Both qualitative and quantitative
query and automatically combines the learned interest to offer modeling techniques can be applied to big data to delve deeper
recommendations (Yin et al. 2014). The performance of the into consumer minds. Econometric modeling with text mining
recommender system depends on the recommendation context techniques can be used for predictive modeling of future
(Huang 2011). For instance, recommendations for high- and changes in sales (Archak et al. 2011). According to Archak et
low-involvement products would vary based on user ratings al. (2011), textual content in product reviews can significantly
and context. The recommender system model integrates user predict future consumer behavior and forecast demand with
ratings and review texts to link latent factors for topic changes in product price, product age, trends, and seasonal
distribution. The aspect-based latent factor model (ALFM) effects. Netzer et al. (2012) compared market structure based on
integrates ratings and review texts via a latent factor model; by user-generated content with sales and survey-based data; results
integrating rating, user-review, and item-attribute matrixes, suggest a significant correlation between the two methods
word latent factors can be derived (Qui et al., 2016). In (Netzer et al. 2012). Main concerns revealed through text
collaborative recommender systems, the overall ratings do not mining match significantly with the main concerns revealed
offer detailed reasoning behind the user's preferences, but through traditional sales and survey-based data (Netzer et al.
124 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

2012). Text mining provides additional associations that need appropriate strategy in future studies. Analysis of market
the attention of marketers. structure based on text mining is more often descriptive and
explains past/present phenomenon. Text mining has enormous
Future Research Directions. potential to predict the future, and future studies can unlock the
predictive modeling potential of data mining (Netzer et al.
Previous research studies of recommender systems are 2012). Incorporation of multiple trails of communications on
context-based. For instance, Huang (2011) studied a recom- different channels, including social media handles, may
mender system for shopping, but the need and effectiveness of improve the predictive ability of the model (Netzer et al.
the recommender system may vary with the nature of the 2012). The only caution is to filter the big data for fitness for
products and services. Convenience and specialty goods use, as online reviews may be biased due to online herd
involve a different level of complexity in consumer decision behavior (Huang and Chen, 2006). It is necessary to check the
making. It would be interesting to understand the comparative representative population to minimize the noise in the findings
performance of a recommender system for different product (Dellarocas, 2006). Data can be full of incorrectly spelled
types (context). Moreover, future studies should attempt to words and unstructured themes, so future researchers should
compare recommender system dynamics for services and goods preprocess data before analysis (Ravi & Ravi 2015).
purchase behaviors. Evolved techniques such as artificial intelligence and
The use of corating data for a similarity index has certain machine learning can provide automated analysis for rich
limitations (Adomavicius & Tuzhilin 2005; Zeng, Xing, Zhou, consumer insights (Ravi & Ravi 2015). Ontology orientation in
& Zheng 2004), so future researchers should incorporate past studies proved successful in representing, visualizing, and
multiattribute utility data for better results. Combined analysis determining sentiment units (Kontopoulos, Berberidis,
of rating and review data in the latent factor model also has Dergiades, & Bassiliades 2013; Lau, Li, & Liao 2014;
constraints (Qiu, Gao, Cheng, & Guo 2016), and incorporation Martinez, Deagustini, Falappa, & Simari 2014). A combined
of additional contextual factors such as geometrical or temporal approach with machine learning and ontological orientation can
information in future studies can boost the predictive ability of open new research opportunities in the future (Ravi & Ravi
the model. Moreover, modeling a recommender system on 2015). Ontology applied in a native language can make it easier
long-horizon longitudinal data will provide strategic implica- to access latent consumer psychology. Currently, opinion
tions for a recommender system for forecasting sales. A mining of non-English language is not available, which limits
combination of prior- and postpurchase recommendation data the reach of emerging technologies. In the future, opening
for model building in future studies will further improve the mapping of cross-linguistic populations can further strengthen
scope of the recommender system. the eWOM research field. Cross-linguistic and cross-domain
Existing research differentiated the vertically differentiated sentiment analysis remains a challenge (Ravi & Ravi 2015).
products such as electronics (Archak et al. 2011), but online Aspect level sentiment analysis can help in understanding the
reviews for horizontally differentiated products are expected to laggard relationship between reviews and sales.
be more complex. Horizontally differentiated products are Emerging techniques such as natural language processing
difficult to evaluate based on quality parameters due to the (NLP), deep learning, and machine learning have already
marginal differences between them. Future researchers should opened new dimensions to the eWOM research field; however,
attempt to explore text mining techniques for horizontally there are additional intelligent techniques ready to emerge
differentiated products to define the small differences perceived shortly. Intelligent techniques such as the random forest,
among consumers. Previous studies assumed that consumers evolutionary computation, association rule mining, fuzzy rule-
learn product attributes linearly, but consumer learning is based systems, rule miner, conditional random field theory
fuzzy. Future studies should attempt to build a more advanced (CRF), formal concept analysis, radial basis function neural
model to incorporate a multidimensional and multiproduct network (RBFNN), and online learning algorithms are emerg-
evaluation environment. ing techniques that can be exploited in the near future. Since
Semantics characteristics in online reviews at the aggregated sentiments are often fuzzy, fuzzy logic offers a tremendous
level provide generalized cues (Cao et al. 2011) and contain opportunity for future researchers. RBFNN and online learning
rich inputs for future decision making. Each component of a algorithms are especially useful for mining data in realtime.
semantic is a cue and to develop a robust model, future studies
should delve into the specific component of semantics. It is Conclusion
important to use multiple single value decomposition (SVD)
from multiple datasets and identify the most significant factors There has been a paradigm shift in marketing communica-
(Sidorova et al.). Existing studies focus on the most frequent tion channels in recent years. Marketing-sponsored communi-
words for analysis. However, rare or infrequent words in the cation channels are being replaced with customer-to-customer
long tail of the distribution may reflect emerging consumer social network channels. Products and services are under the
preferences and could be very helpful in developing a new stringent vigil of consumers. The consumers are sharing their
marketing model (Guo et al. 2017). product and service experiences in virtual communities that
The use of text mining tools during different stages of the have no geographical boundary. Moreover, such electronic
product lifecycle can become instrumental in devising word of mouth is more prevalent, relevant, and needed
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 125

specifically in the case of service due to its intangible become transversal themes. The recommender system and data
characteristics. Service is more experience dominated and mining became the motor themes, showing their growing
difficult to evaluate before consumption. Therefore, consumers popularity. The stability index for a recommender system is
rely on shared experiences and eWOM to reduce the perceived 0.65, user-generated content is 0.58, social networking is 0.48,
risk associated with services. Thus, research on eWOM became and data mining is 0.42. This shows that these themes have a
popular and attracted the attention of researchers globally. consistent evolution in the eWOM research domain.
Several studies tried to understand the antecedents and Further, we used cocitation analysis for mapping the
consequences of eWOM. Due to wider scope and extensive intellectual structure of the research field. Cocitation analysis
coverage, the metasynthesis of existing studies for identifica- resulted in a three-cluster structure. Studies in cluster 1 address
tion of future research directions is extremely important. Given the conceptualization of eWOM, studies in cluster 2 focused on
this, previous researchers have attempted to author systematic the impact of eWOM on sales, and studies in cluster 3 revolve
literature reviews, but existing reviews are mostly descriptive, around the implication of eWOM communication on experi-
and the latent conceptual and intellectual structure was ence dominated sectors such as hospitality and tourism.
unexplored. To fill this research gap, the present study attempts Previous researchers used a multitude of methodological
a comprehensive performance analysis of the research domain techniques such as log-linear modeling (Ye et al., 2009) and
and science mapping with conceptual and intellectual structure experimentation (Vermeulin and Seegers, 2009) to understand
analyses. the impact of eWOM.
A scientifically established literature review methodology, Content analysis of emerging topics and emerging themes
as suggested by Saunders et al. (2009), Rowley and Slack helped in the identification of ongoing debates and future
(2004), and Fahimnia et al. (2015), was followed to conduct research directions. Recommender system research may see
this study. The performance analysis of different scientific shifts in the research lens and is expected to offer effective
actors offers the structure and development of the eWOM predictive modeling. Due to the complexity in corating data,
research field. Various scientific actors used for the analysis are future researchers should incorporate multiattribute utility data
journals, articles, authors, and keywords. Both quantitative and for better results. Incorporation of additional factors such as
qualitative aspects of performance analysis provided vital geometrical or temporal information along with contextual data
insights for future researchers. Quantitative measures such as in future studies can boost the predictive ability of the models.
frequency of publications and qualitative measures such as Evolved techniques such as artificial intelligence and machine
citation index are used for performance analysis. Finally, learning can provide automated analysis for rich consumer
cooccurrence and cocitation analyses helped to understand the insights (Ravi & Ravi 2015). A combined approach with
conceptual and intellectual structure of the existing literature machine learning and ontological orientation can open new
for the identification of future research directions. research avenues in the future (Ravi & Ravi 2015). In the
In this study, we reviewed 1896 research papers published future, opinion mining of cross-linguistic populations can
during the last two decades (between 2000 and 2020). further strengthen the eWOM research field.
Longitudinal data analysis indicates the growing popularity of This study makes a significant theoretical contribution to the
the research field. We used coword analysis to map the eWOM research field. It extended the existing knowledge by
conceptual structure of the eWOM research. Recommender offering performance analysis and science mapping (conceptual
systems and data mining emerged as the most promising topics and intellectual structures) of the research field. This study
within the eWOM research field. Data mining includes went beyond a mere systematic literature review and offered an
keywords such as text mining, semantic analysis, and sentiment indepth content analysis of emergent themes for identification
analysis. It clearly shows the usage of both quantitative and of future research directions. This study also offers insights for
qualitative approaches for mining eWOM information. The practitioners to leverage the vast potential of eWOM for brand
recommender system addressed specific issues such as online building. The systematic presentation of existing literature will
rating and collaborative filtering for predictive modeling. help practitioners link different antecedents and consequences
We plotted a thematic evolution map (alluvial diagram) to of eWOM. The understanding of emerging themes and topics
understand the longitudinal progression of the research field. will help practitioners devise future strategies accordingly.
Themes such as decision making and recommender system Interwoven technology intervention in realtime analysis of big
evolve continuously, and the diagram shows their continually data is expected to create a new paradigm for both
growing importance. Some thematic areas, such as the Internet, academicians and practitioners.
ecommerce, and user-generated concepts, showed a sinusoidal
pattern. For a microscopic understanding of eWOM research References
evolution, we plotted the strategic map of each subperiod to
understand the importance of growing themes by phase. Adomavicius, G., & Tuzhilin, A. (2005). Toward the next generation of
Initially, the Internet garnered significant attention, and various recommender systems: a survey of the state-of-the-art and possible
studies attributed the increase in eWOM to the growth of the extensions. IEEE Transactions on Knowledge and Data Engineering, 17
(6), 734–749.
Internet. With time, online social networking became the Alonso, S., Cabrerizo, F., Herrera-Viedma, E., & Herrera, F. (2009). H-index: a
central theme. In recent years, social media became the basic review focused in its variants, computation and standardization for different
theme of studies, while eWOM and user-generated content scientific fields. Journal of Informetrics, 3, 273–289.
126 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

Archak, N., Ghose, A., & Ipeirotis, P. G. (2011). Deriving the pricing power of Culnan, M. (1986). The intellectual development of management information
product features by mining consumer reviews. Management Science, 57(8), systems. Management Science, 32(2), 156–172.
1485–1509. Davis, A., & Khazanchi, D. (2008). An empirical study of online word of mouth
Arndt, J. (1967). Role of product-related conversations in the diffusion of a new as a predictor for multi-product category e-commerce sales. Electronic
product. Journal of Marketing Research, 4(3), 291. Markets, 18(2), 130–141.
Barreto, A. M. (2015). The word-of-mouth phenomenon in the social media era. Day, G. S. (1971). Attitude change, media and word of mouth. Journal of
International Journal of Market Research, 56(5), 431. Advertising Research, 11(6), 31–40.
Baum, D., & Spann, M. (2014). The interplay between online consumer reviews Dellarocas, C. (2003). The digitization of word of mouth: promise and
and recommender systems: an experimental analysis. International Journal challenges of online feedback mechanisms. Management Science, 49(10),
of Electronic Commerce, 19(1), 129–162. 1407–1424.
Baur, T., & Nyström, B. (2017). The effect of eWOM components on brand Duan, W., Gu, B., & Whinston, A. B. (2008). Do online reviews matter? An
attitude and purchase intention: A cross country study between Germany empirical investigation of panel data. Decision Support Systems, 45(4),
and Sweden. 1007–1016.
Berezina, K., Bilgihan, A., Cobanoglu, C., & Okumus, F. (2016). Understand- Fahimnia, B., Sarkis, J., & Davarzani, H. (2015). Green supply chain
ing satisfied and dissatisfied hotel customers: text mining of online hotel management: a review and bibliometric analysis. International Journal of
reviews. Journal of Hospitality Marketing & Management, 25(1), 1–24. Production Economics, 162, 101–114.
Bickart, B., & Schindler, R. M. (2001). Internet forums as influential sources of Forman, C., Ghose, A., & Wiesenfeld, B. (2008). Examining the relationship
consumer information. Journal of Interactive Marketing, 15(3), 31–40. between reviews and sales: the role of reviewer identity disclosure in
Blondel, V. D., Guillaume, J. L., Lambiotte, R., & Lefebvre, E. (2008). Fast electronic markets. Information Systems Research, 19(3), 291–313.
unfolding of communities in large networks. Journal of Statistical Garfield, E. (1972). Citation analysis as a tool in journal evaluation. Science,
Mechanics: Theory and Experiment. 178, 471–479.
Börner, K., Chen, C., & Boyack, K. (2003). Visualizing knowledge domains. Garfield, E. (1994). Scientography: mapping the tracks of science. Current
Annual Review of Information Science and Technology, 37, 179–255. Contents: Social & Behavioural Sciences, 7, 5–10.
Braam, R. R., Moed, H. F., & van Raan, A. F. J. (1991). Mapping of science by Guo, Y., Barnes, S. J., & Jia, Q. (2017). Mining meaning from online ratings
combined co-citation and word analysis. II: Dynamical aspects. Journal of and reviews: tourist satisfaction analysis using latent dirichlet allocation.
the American Society for Information Science, 42, 252–266. Tourism Management, 59, 467–483.
Brown, J., Broderick, A. J., & Lee, N. (2007). Word of mouth communication Gupta, S., & Kim, H. W. (2004). Virtual community: concepts, implications,
within online communities: conceptualizing the online social network. and future research directions. In Proceedings of the tenth Americas
Journal of Interactive Marketing, 21(3), 2–20. conference on information systems (Vol. 10, pp. 2679-87).
Cabrerizo, F. J., Alonso, S., Herrera-Viedma, E., & Herrera, F. (2010). q2- Hao, Y., Ye, Q., Li, Y., & Cheng, Z. (2010). How does the valence of online
index: quantitative and qualitative evaluation based on the number and consumer reviews matter in consumer decision making? Differences
impact of papers in the Hirsch core. Journal of Informetrics, 4, 23–28. between search goods and experience goods. In 2010 43rd Hawaii
Cahlik, T. (2000). Comparison of the maps of science. Scientometrics, 49, International Conference on System Sciences (pp. 1-10). IEEE.
373–387. Henderson, M., Shurville, S., & Fernstrom, K. (2009). The quantitative crunch:
Callon, M., Courtial, J. P., & Laville, F. (1991). Co-word analysis as a tool for the impact of bibliometric research quality assessment exercises on
describing the network of interactions between basic and technological academic development at small conferences. Campus-Wide Information
research—the case of polymer chemistry. Scientometrics, 22, 155–205. Systems, 26, 149–167.
Callon, M., Courtial, J. P., Turner, W. A., & Bauin, S. (1983). From translations Hennig-Thurau, T., Gwinner, K. P., Walsh, G., & Gremler, D. D. (2004).
to problematic networks: an introduction to co-word analysis. Social Electronic word-of-mouth via consumer-opinion platforms: what motivates
Science Information, 22, 191–235. consumers to articulate themselves on the internet? Journal of Interactive
Cantallops, A. S., & Salvi, F. (2014). New consumer behavior: a review of Marketing, 18(1), 38–52.
research on eWOM and hotels. International Journal of Hospitality Hirsch, J. (2005). An index to quantify an individual's scientific research
Management, 36, 41–51. output. Proceedings of the National Academy of Sciences, 102, 16569–
Cao, Q., Duan, W., & Gan, Q. (2011). Exploring determinants of voting for the 16572.
“helpfulness” of online user reviews: a text mining approach. Decision Hjørland, B. (2013). Citation analysis: asocial and dynamic approach to
Support Systems, 50(2), 511–521. knowledge organization. Information Processing Management, 49(6),
Chan, Y. Y., & Ngai, E. W. (2011). Conceptualising electronic word of mouth 1313–1325.
activity (Marketing Intelligence & Planning). Huang, S. L. (2011). Designing utility-based recommender systems for e-
Charlett, D., Garland, R., & Marr, N. (1995). How damaging is negative word commerce: evaluation of preference-elicitation methods. Electronic Com-
of mouth. Marketing Bulletin, 42–50. merce Research and Applications, 10(4), 398–407.
Chen, C., Ibekwe-San Juan, F., & Hou, J. (2010). The structure and dynamics of Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The
co-citation clusters: a multiple perspective co-citation analysis. Journal of challenges and opportunities of social media. Business Horizons, 53(1),
the American Society for Information Science, 61(7), 1386–1409. 59–68.
Cheung, M. Y., Luo, C., Sia, C. L., & Chen, H. (2009). Credibility of electronic Katz, E., & Lazarsfeld, P. F. (1955). Personal influence. New York, 792.
word-of-mouth: informational and normative determinants of on-line Keller, K. L. (2007). Strategic brand management: Building, measuring, and
consumer recommendations. International Journal of Electronic Com- managing brand equity 3rd ed. New York: Prentice Hall.
merce, 13(4), 9–38. Kontopoulos, E., Berberidis, C., Dergiades, T., & Bassiliades, N. (2013).
Cheung, C. M. K., & Thadani, D. R. (2012). The impact of electronic word-of- Ontology-based sentiment analysis of twitter posts. Expert Systems with
mouth communication: a literature analysis and integrative model. Decision Applications, 40(10), 4065–4074.
Support Systems, 54(1), 461–470. Krsul, I. (1998). Software vulnerability analysis. Ph.D. thesis. Purdue
Chevalier, J. A., & Mayzlin, D. (2006). The effect of word of mouth on sales: University.
online book reviews. Journal of Marketing Research, 43(3), 345–354. Lau, R. Y., Li, C., & Liao, S. S. (2014). Social analytics: learning fuzzy product
Clauset, A., Newman, M. E. J., & Moore, C. (2004). Finding community ontologies for aspect-oriented sentiment analysis. Decision Support
structure in very large networks. Physical Review, 70(6), 1–6. Systems, 65, 80–94.
Coulter, N., Monarch, I., & Konda, S. (1998). Software engineering as seen Lee, M., Rodgers, S., & Kim, M. (2009). Effects of valence and extremity of
through its research literature: a study in co-word analysis. Journal of the eWOM on attitude toward the brand and website. Journal of Current Issues
American Society for Information Science, 49, 1206–1223. & Research in Advertising, 31(2), 1–11.
S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128 127

Lee, M., & Youn, S. (2009). Electronic word of mouth (eWOM) how eWOM Saunders, M. L., & Lewis, P. (2009). P. & Thornhill, A. (2009). Research
platforms influence consumer product judgement. International Journal of methods for business students, Pearson Education.
Advertising, 28(3), 473–499. Scott, S. G., & Lane, V. R. (2000). A stakeholder approach to organizational
Leydesdorff, L. (2011). Bibliometrics/citation networks. In G.A. Barnett (Ed.) identity. Academy of Management Review, 25(1), 43–62.
Encyclopaedia of social networks, Thousand Oaks, CA: SAGE Publications Sharplin, A., & Mabry, R. (1985). The relative importance of journals used in
Inc.. management research: an alternative ranking. Human Relations, 38(2),
Li, X., & Wang, Y. (2011). China in the eyes of western travelers as represented 139–149.
in travel blogs. Journal of Travel & Tourism Marketing, 28(7), 689–719. Sidorova, A., Evangelopoulos, N., Valacich, J. S., & Ramakrishnan, T. (2008).
Litvin, S. W., Goldsmith, R. E., & Pan, B. (2008). Electronic word-of-mouth in Uncovering the intellectual core of the information systems discipline. MIS
hospitality and tourism management. Tourism Management, 29(3), Quarterly, 32(3), 467–482.
458–468. Small, H. (1973). Co-citation in the scientific literature: a new measure of the
Liu, Y. (2006). Word of mouth for movies: its dynamics and impact on box relationship between two documents. Journal of the American Society for
office revenue. Journal of Marketing, 70(3), 74–89. Information Science, 24(4), 265–269.
Mac Roberts, M. H., & Mac Roberts, B. R. (1989). Problems of citation Small, H. (1977). A co-citation model of a scientific specialty: a longitudinal
analysis: a critical review. Journal of the American Society for Information study of collagen research. Social Studies of Science, 7, 139–166.
Science, 40(5), 342–349. Small, H. (1999). Visualizing science by citation mapping. Journal of the
Martinez, M. V., Deagustini, C. A. D., Falappa, M. A., & Simari, G. R. (2014, American Society for Information Science, 50, 799–813.
November). Inconsistency-tolerant reasoning in data log ontologies via an Small, H. (2006). Tracking and predicting growth areas in science.
argumentative semantics. In Ibero-American Conference on Artificial Scientometrics, 68, 595–610.
Intelligence (pp. 15-27). Springer, Cham. Sparks, B. A., & Browning, V. (2011). The impact of online reviews on hotel
Mishra, D., Gunasekaran, A., Childe, S., & Papadopoulos., & Dubey, R. booking intentions and perception of trust. Tourism Management, 32(6),
(2016). Vision, applications and future challenges of internet of things: a 1310–1323.
bibliometric study of the recent literature. Industrial Management and Data Tirunillai, S., & Tellis, G. J. (2014). Mining marketing meaning from online
Systems, 116(7), 1331–1355. chatter: strategic brand analysis of big data using latent dirichlet allocation.
Mishra, A., & Satish, S. M. (2016). eWOM: extant research review and future Journal of Marketing Research, 51(4), 463–479.
research avenues. Vikalpa, 41(3), 222–233. Tranfield, D., Denyer, D., & Smart, P. (2003). Towards a methodology for
Mudambi, S. M., & Schuff, D. (2010). What makes a helpful online review? A developing evidence-informed management knowledge by means of
study of customer reviews on Amazon. Com. MIS Quarterly, 34(1), systematic review. British Journal of Management, 14(3), 207–222.
185–200. Tuma, M. N., & Decker, R. (2013). Online reviews as a source of marketing
Netzer, O., Feldman, R., Goldenberg, J., & Fresko, M. (2012). Mine your own research data: a literature analysis. International Journal of Electronic
business: market-structure surveillance through text mining. Marketing Marketing and Retailing, 5(4), 287–316.
Science, 31(3), 521–543. Van Raan, A. F. J. (2005). Handbook of quantitative science and technology
Noyons, E. C. M., Moed, H. F., & van Raan, A. F. J. (1999). Integrating research. Chapter measuring science. Netherlands: springer., pp. 19–50.
research performance analysis and science mapping. Scientometrics, 46, Varadarajan, P. R., & Yadav, M. S. (2002). Marketing strategy and the internet:
591–604. an organizing framework. Journal of the Academy of Marketing Science, 30
Palanivel, K., & Sivakumar, R. (2011). A study on collaborative recommender (4), 296–312.
system using fuzzy-multicriteria approaches. International Journal of Vázquez-Casielles, R., Suárez-Álvarez, L., & del Río-Lanza, A. B. (2013). The
Business Information Systems, 7(4), 419–439. word of mouth dynamic: how positive (and negative) WOM drives purchase
Park, D. H., Lee, J., & Han, I. (2007). The effect of on-line consumer reviews probability: an analysis of interpersonal and non-interpersonal factors.
on consumer purchasing intention: the moderating role of involvement. Journal of Advertising Research, 53(1), 43–60.
International Journal of Electronic Commerce, 11(4), 125–148. Vermeulen, I. E., & Seegers, D. (2009). Tried and tested: the impact of online
Pilkington, A., & Catherine, L. H. (1999). Is production and operations hotel reviews on consumer consideration. Tourism Management, 30(1),
management a discipline? A citation/co-citation study. International 123–127.
Journal of Operations and Production Management, 19(1), 7–20. Vokurka, R. J. (1996). The relative importance of journals used in operations
Purnawirawan, N., Eisend, M., De Pelsmacker, P., & Dens, N. (2015). A meta- management research: a citation analysis. Journal of Operations Manage-
analytic investigation of the role of valence in online reviews. Journal of ment, 14(4), 345–355.
Interactive Marketing, 31, 17–27. Weisfeld-Spolter, Suri, Sussan, Fiona, & Gould, Stephen (2014). An integrative
Qiu, L., Gao, S., Cheng, W., & Guo, J. (2016). Aspect-based latent factor model approach to eWOM and marketing communications. Corporate Communi-
by integrating ratings and reviews for recommender system. Knowledge- cations: An International Journal, 19(3), 260–274.
Based Systems, 110, 233–243. Westbrook, R. A. (1987). Product/consumption-based affective responses and
Radicchi, F., Castellano, C., Cecconi, F., Loreto, V., & Parisi, D. (2004). post purchase processes. Journal of marketing research, 258-270.
Defining and identifying communities in networks. Proceedings of the Xia, L., & Bechwati, N. N. (2010). Word of mouse: the role of cognitive
National Academy of Sciences of the United States of America, 101(9), personalization in online consumer reviews. Journal of Interactive
2658–2663. Advertising, 9(1), 3–13.
Ramos-Rodríguez, A. R., & Ruíz-Navarro, J. (2004). Changes in the intellectual Xiang, Z., & Gretzel, U. (2010). Role of social media in online travel
structure of strategic management research: a bibliometric study of the information search. Tourism Management, 31(2), 179–188.
strategic management journal, 1980–2000. Strategic Management Journal, Ye, Q., Law, R., Gu, B., & Chen, W. (2011). The influence of user-generated
25(10), 981–1004. content on traveler behavior: an empirical investigation on the effects of e-
Ravi, K., & Ravi, V. (2015). A survey on opinion mining and sentiment word-of-mouth to hotel online bookings. Computers in Human Behavior,
analysis: tasks, approaches and applications. Knowledge-Based Systems, 89, 27(2), 634–639.
14–46. Yin, H., Cui, B., Sun, Y., Hu, Z., & Chen, L. (2014). LCARS: a spatial item
Rayport, J. F., & Sviokla, J. J. (1994). Managing in the marketspace. Harvard recommender system. ACM Transactions on Information Systems (TOIS),
Business Review, 72(6), 141–150. 32(3), 1–37.
Rosvall, M., & Bergstrom, C. T. (2010). Mapping change in large networks. Yong-Hak, J. (2013). Web of science (Thomson Reuters).
PLoS One, 5. Zarrad, H., & Debabi, M. (2015). Analysing the effect of electronic word of
Rowley, J., & Slack, F. (2004). Conducting a literature review. Management mouth on tourists' attitude toward destination and travel intention.
Research News, 27, 31–39. International research journal of social sciences, 4(4), 53–60.
128 S. Verma, N. Yadav / Journal of Interactive Marketing 53 (2021) 111–128

Zeng, C., Xing, C.-X., Zhou, L.-Z., & Zheng, X.-H. (2004). Similarity measure adherence research. Journal of the Association for Information Science and
and instance selection for collaborative filtering. International Journal of Technology, 67(4), 967–972.
Electronic Commerce, 8(4), 115–129. Zhu, F., & Zhang, X. (2010). Impact of online consumer reviews on sales: the
Zhang, J., Yu, Q., Zheng, F., Long, C., Lu, Z., & Duan, Z. (2016). Comparing moderating role of product and consumer characteristics. Journal of
keywords plus of WOS and author keywords: a case study of patient Marketing, 74(2), 133–148.

You might also like