See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/297742527

Twitter sentiment analysis: Capturing sentiment from integrated resort
tweets
Article in International Journal of Hospitality Management · May 2016
DOI: 10.1016/j.ijhm.2016.02.001

CITATIONS

READS

3,672

2 authors:
Kahlil S. Philander

Yunying Susan Zhong

Washington State University

Rosen College of Hospitality Management, University of Central Florida

29 PUBLICATIONS 215 CITATIONS

11 PUBLICATIONS 158 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Casino Loyalty View project

All content following this page was uploaded by Kahlil S. Philander on 06 March 2019.
The user has requested enhancement of the downloaded file.

SEE PROFILE

International Journal of Hospitality Management 55 (2016) 16–24

Contents lists available at ScienceDirect

International Journal of Hospitality Management
journal homepage: www.elsevier.com/locate/ijhosman

Twitter sentiment analysis: Capturing sentiment from integrated
resort tweets
Kahlil Philander a,1 , YunYing Zhong b,∗
a

William F. Harrah College of Hotel Administration, University of Nevada, Las Vegas, 4505 S. Maryland Parkway, Las Vegas, NV 89154, United States
Hospitality Management, College of Health and Human Services, California State University, Long Beach, 1250 Bellﬂower Blvd, Long Beach, CA 90840,
United States
b

a r t i c l e

i n f o

Article history:
Received 10 October 2014
Received in revised form 26 January 2016
Accepted 2 February 2016
Available online 11 March 2016

1. Introduction
With the rapid development of information and communication technologies, social networking sites and blogs have become
an enormous repository for rich user generated content (UGC)
(Thelwall et al., 2011). Information system scholars observed that
business intelligence has evolved from analysis of traditionally
structured business transaction data to studying unstructured and
real-time data, driven by the development of online search engines,
e-commerce, and social media (Chen et al., 2012). Among the existing collection of social media platforms, the Twitter microblog is
one of the most popular mediums that have been adopted by both
consumers and companies. Over half a billion Twitter messages
(tweets) per day are being recorded (Kirilenko and Stepchenkova,
2014; Krikorian, 2013), and a recent customer engagement technology study reported that Twitter is among the top three most
used social media platforms by US hotel and restaurant operators
(Kim and Connolly, 2013).
The prevalence of Twitter and its prodigious amount of UGC
creates important implications for the hospitality and tourism sector. Research has shown that online UGC has become an important
information source that exerts critical inﬂuence on customers’
brand perception, brand reputation, purchase decision making, and
proﬁtability (Browning et al., 2013; Leung et al., 2013; Vermeulen
and Seegers, 2009; Ye et al., 2009, 2011; Zhang et al., 2010). Twitter
may be particularly susceptible to the effects of electronic words
of mouth (EWOM) due to its viral nature. According to Kwak et al.

∗ Corresponding author. Tel.: +1562 985 4496.
E-mail addresses: kahlil.philander@unlv.edu (K. Philander),
yunying.zhong@csulb.edu (Y. Zhong).
Tel.: +1 702 895 1265.
http://dx.doi.org/10.1016/j.ijhm.2016.02.001
0278-4319/© 2016 Elsevier Ltd. All rights reserved.

(2010) study, a tweet, if resent by a different user (retweeted), is
expected to reach an average of 1000 registered users. In three years
following Kwak et al.’s study, the total volume of daily tweets has
grown over ten-fold (Krikorian, 2013; Weil, 2010).
While the viral inﬂuence of Twitter messages makes it valuable
for ﬁrms to monitor and identify customer comments, the massive volume and high variance of real-time tweets has made that
task extremely time consuming, costly, and often manually impossible (Chiu et al., 2015; Claster et al., 2013). Even though ﬁrms have
started to embrace Twitter as a marketing platform, because of
these obstacles, they often fail to extract much objective quantitative feedback or establish competitive benchmarks (i.e. comparing
ﬁrm performance in social media against competitor performance)
from this available data.
Sentiment analysis, which is still a rather nascent arena, appears
to be a promising tool in solving the above mentioned problem.
This technique involves using algorithms to automatically extract
and classify text data into sentiment categories: positive, negative
and/or neutral (Chiu et al., 2015; Pang and Lee, 2008). As compared to more traditional market research methods (e.g. surveys
or opinion polls), sentiment analysis not only has the advantage
of being more cost and time efﬁcient in many cases, it is a nonintrusive method to extract consumers’ opinions and sentiments in
“real-time”—avoiding recall biases generally associated with postconsumption self-report measurements (Rylander et al., 1995).
Furthermore, sentiment analysis can provide a temporal sentiment
proﬁle even on a second by second scale, which is not generally
possible for survey based market analysis to achieve.
In this study, we demonstrate the application of sentiment analysis using Twitter data to build low-cost and real-time measures
of hospitality customer attitudes/perceptions. Using a popular
tourism destination (Las Vegas, NV) as a case study, we create a sentiment index for every Twitter account belonging to an

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

integrated-resort property in the Las Vegas metropolitan area. The
ensuing sentiment metrics are then used to benchmark these ﬁrms
against one another, and compare their performance over time. The
metrics are also cross checked with external hotel-casino rankings
(i.e. TripAdvisor) to establish external validity. Discussion of the
usefulness of this business intelligence tool is provided, along with
potential limitations.

2. Literature review
2.1. Twitter and its applications
Since its inception in 2006, Twitter has become one of the most
inﬂuential microblogging platforms on the Internet (Akehurst,
2009; Thelwall et al., 2011). According to a recent article on Hufﬁngton Post, Twitter ofﬁcially reported a total number of 218.3 million
“monthly active users,” or accounts in its 2013 October ﬁling with
the Security and Exchange Commission (Grandoni, 2013). In the
United States alone, Nielsen (2012) estimated that in 2012, the
microblogging site had over 37 million users via personal computer access, and about another 65 million via mobile devices. In
addition to its large user base, Twitter users show a high level of
engagement. Nielsen (2012) included Twitter among the top-three
sites in terms of users’ total time spent (with Facebook and Tumblr).
A recent study conducted by the Pew Research Center also found
that 46% of Twitter users tweet on a daily basis (Dugann and Smith,
2013).
With its large user base and high engagement, Twitter has
increasingly been used by the hospitality and tourism industry
for promotion, distribution, marketing management, communication, and market research (Leung et al., 2013). When studying
the use of social media by the hotel industry in Hong Kong, Chan
and Guillet (2011) found that Twitter and Facebook were the two
most widely used social sites. The industry mainly used the sites
to promote discounted products and services, answering guests’
inquiries, and handling complaints. To a lesser extent, the sites were
also employed to engage guests and obtain business intelligence.
Twitter is also a popular social medium used by national destination
marketing organizations (DMOs) for marketing purpose. Examining the social media activities of eight top international tourism
destinations, Hays et al. (2013) uncovered that the DMOs generally posted more frequently on Twitter than on Facebook. The
researchers explained that Twitter structure was built for timely
updates and massive information broadcast. Due to the constant
updates on this platform, tweets by any given users can be rapidly
buried. As a result, organizations need to constantly refresh its
updates to remain visible.
In the area of market research, Twitter data tends to be utilized
more than those from other social media platforms. This is in part
because of Twitter’s more liberal data availability; Facebook messages are often not publicly accessible due to its stricter privacy
settings, and other platforms generally have smaller user pools
(Wang et al., 2013). The high utility of Twitter data for research
can also be attributed to its “people as sensor” network structure.
Twitter has a fundamentally open networking structure, which
allows its users to choose who they want to follow without seeking permission (Weng et al., 2010). This open structure of Twitter
has made it into a “global pipeline for real time information sharing and broadcasting” (Wang et al., 2011, p. 34). The interactions
among Twitter users can be viewed as a large network of sensors
that react to external and social events, thus making it particularly
suitable for studying the public opinions of these events (Kirilenko
and Stepchenkova, 2014).
Aside from its conceptualization as a sensor network, Twitter
is conducive for market research due to the internal architecture

of the messages. As a microblogging site, Twitter allows its users
to post messages of a maximum 140 characters in length. These
compact messages, also called “tweets”, capture the primary meaning users want to express without much irrelevant content. The
unique function of hashtags, “#”, enables users to highlight the
key words in their messages or ﬁnd other posts under certain key
words. Tweets can be forwarded (or “retweeted”) into a social network of friends and contacts by the use of “RT<@username>” or via
“<@username>”. The “@” symbol is used to identify tweets which are
addressed to a particular registered user (Cha et al., 2010). The condensed structure of tweets and Twitter’s unique functions make its
data a useful ﬁt for content analysis and sentiment analysis, which
involves systematic reduction of content ﬂow for the purpose of
quantitative analysis (Kirilenko and Stepchenkova, 2014).
Twitter has the advantage of capturing a rich amount of authentic customers’ in-the-moment experience and sentiment (Capriello
et al., 2011; Dodds et al., 2011). As compared to traditional research
methods, such as surveys and comment cards, Twitter is a nonintrusive research model that suffers less from bias associated
with recall and interactions with human subjects (Kirilenko and
Stepchenkova, 2014). Asur and Huberman (2010) note that the collective wisdom on Twitter, if analyzed properly in a large enough
volume, can be more accurate than other techniques in extracting
information, such as surveys and opinion polls. Of course, Twitter
data has limitations. For example, Twitter users are not representative of the general population, and Twitter posters are not
necessarily representative of the entire Twitter user base–Twitter
Inc. noted that 40% of users do not post any content. Boyd and
Crawford (2012) caution that researchers should understand, and
publicly account for, the limits of Twitter data set analyses.
2.2. Electronic word of mouth and its impacts
The attention social media receive from hospitality and tourism
practitioners not only can be attributed to its value for business
intelligence, but also to its powerful ability to spread word of mouth
(WOM) immediately on a large scale (Leung et al., 2013). WOM is
deﬁned as informal communications among consumers concerning
a product, service or an organization, and it is distinguished from
those communications in which companies pass product knowledge to consumers through massive media (Litvin et al., 2008).
Research has long recognized WOM as a trustworthy information
source independent from commercial inﬂuences perceived by consumers (Leung et al., 2013). Traditional WOM and EWOM share a
lot of similarity. However, the latter has the ability to reach beyond
one’s immediate social circle and transcend the ephemeral nature
of ofﬂine WOM (Litvin et al., 2008). EWOM, often in the form of
online reviews and comments, can be easily archived, linked, and
searched. EWOM is increasingly important in shaping consumers’
product knowledge and perceptions (Litvin et al., 2008)
A thorough literature review on hospitality and tourism EMOWonline review was conducted by Schuckert et al. (2015). Examining
50 relevant articles between 2004 and 2013, the researchers summarized ﬁve thematic clusters within the literature: the effect of
online review on consumers’ buying, online response management
and consumer satisfaction, reviewer’s motivation, sentiment analysis, and critical attributes of online review. The researchers noted
more attention paid to the ﬁrst two consumer-concentric topics
and more studies (60%) done in the hotel setting. EWOM is also
naturally integral with the social media literatures. After a comprehensive review on social media in tourism and hospitality, Leung
et al. (2013) pointed out that EWOM serves as a critical source for
consumer’s pre-trip travel planning, and can be generated throughout their trip and post-trip phase.
As such, prior hospitality and tourism EWOM research can also
be classiﬁed in two perspectives, listeners’ and originators’ (Litvin

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

et al., 2008). The studies from the listener viewpoint mainly focus
on the effect of EWOM on consumers’ attitudes and behavioral
intention during the planning phase. As Leung et al. (2013) noted,
because tourism and hospitality related products are intangible and
often costly, consumers tend to seek others’ opinions to reduce risks
associated with their decision making. Studies have shown that the
valence of the EWOM has signiﬁcant impacts on consumers’ trust
(Sparks and Browning, 2011; Ladhari and Michaud, 2015), brand
awareness (Vermeulen and Seegers, 2009), and purchase intentions (Mauri and Minazzi, 2013; Vermeulen and Seegers, 2009). The
number of reviews and the numerical rating also greatly affect consumers’ perception, webpage visits, and booking behaviors (Sparks
and Browning, 2011; Ye et al., 2009, 2011; Zhang et al., 2010).
Furthermore, other information cues, such as source of reviews,
writing style, and reviewer’s expertise and personal identiﬁable
information, can play a signiﬁcant role in affecting consumer’s decision making (Sparks et al., 2013; Xie et al., 2011). Based on these
observed contributions of EWOM to critical marketing outcomes,
Ghose and Ipeirotis (2009) posited that each snippet of UGC has
economic value, and organizations with more favorable consumer
evaluation can charge a price premium.
From the originator’s perspective, many studies focus on
consumers’ motivations for originating and spreading WOM.
Hennig-Thurau et al. (2004) found that the primary reasons for
consumers to speak on the Internet are their desire for social
interaction, economic incentives, concern for other consumers, and
potential to enhance self-worth. Similarly, Litvin et al. (2008) summarized the ﬁndings of previous research, and suggested that the
motivations of consumers’ WOM behaviors include expressing positive feeling, discharging negative emotions, conducting altruistic
acts, reciprocating with others, and enjoying shared experiences.
Consumers’ active sharing can drastically change the course or
meaning of marketing messages that are set forth by the companies.
It is important for ﬁrms to closely monitor their social communications. This includes both the volume and the content, which enables
companies to quickly identify trends or crises (Hennig-Thurau et al.,
2010, 2013). Litvin et al. (2008) found that hospitality and tourism
companies can take advantage of EWOM sharing to create “buzz”
which ampliﬁes their initial marketing efforts. This is achieved by
creating or advertising something interesting that can get consumers talking. Research has supported the power of “buzz” with
evidence suggesting that exposure to online comments enhance
consumers’ brand awareness (Vermeulen and Seegers, 2009).
From the listeners’ perspective EWOM exerts substantial inﬂuences on attitudes, behaviors, and purchase intentions. From the
originators’ and sharers’ perspectives, hospitality and tourism
organizations can take advantage of consumer’s EWOM to build
marketing momentum centered on initial marketing activities.
Generally, more positive EWOM has a more positive impact on
consumers (Sparks and Browning, 2011; Vermeulen and Seegers,
2009; Ye et al., 2011). Sentiment analysis can be an effective tool
for measuring these outcomes.
2.3. Sentiment analysis
To extract consumers’ opinions and sentiments from a pool
of qualitative data is not a novel task for marketing researchers.
Traditionally, such a task can be performed by manual content
coding. Researchers examine the data thoroughly and comprehensively, and develop their own sentiment categories independently.
Inter-rater reliability is then calculated to determine the degree
of homogeneity in their classiﬁcations. Divergent ﬁndings are discussed as an attempt to resolve disagreement (Gibbs, 2007). While
manual analysis can better detect nuances in the text (e.g. sarcasm
or implicit sentiments), it is highly time consuming and labor intensive. With the rapidly increasing volume of online data, manual

analysis becomes pragmatically challenging, if not impossible. This
gives strong impetus to the rise of a new area of research, sentiment
analysis.
Sentiment analysis, also referred to as opinion mining, is a technique developed in the ﬁelds of artiﬁcial intelligence and natural
language processing. It is an information retrieval tool that can classify text into subjective categories (negative, positive, or neutral) or
measure sentiment strength (Pang and Lee, 2008; Thelwall et al.,
2010). There are two major steps in sentiment analysis: opinion
extraction and sentiment classiﬁcation (Pang and Lee, 2008). Opinion extraction is to differentiate subjective texts from factual ones,
while sentiment classiﬁcation focuses on assigning opinion words
into different sentiment categories (Chiu et al., 2015). Opinion words
are words that express desirable (e.g. fantastic, amazing, etc.) and
undesirable (terrible, disgusting, etc.) states (Ding et al., 2008).
There are two common methods determining an opinion
word’s semantic orientation or subjective categories: corpus based
and dictionary-based approach (Chiu et al., 2015). Corpus-based
approaches involve using the syntactic and co-occurrence pattern
of a large corpus (texts that are most representative of a document’s
content) in identifying sentiment category (Capriello et al., 2011;
Chiu et al., 2015; Thelwall et al., 2011). For instance, Turney and
Littman (2003) determined a word’s semantic orientation by calculating the strength of its association with a set of positive words
minus its association with a set of negative words. The associations were estimated by issuing a search engine query, and then
noting the query’s co-occurrence probability with the positive and
negative seed words.
A dictionary-based approach, also called a lexicon-based
method, uses a bank of pre-coded words to determine the text’s
semantic orientation (Chiu et al., 2015; Thelwall et al., 2011). For
example, Hu and Liu (2004) generated a set of adjective synonyms
and antonyms (opinion words) through bootstrapping process
using the WordNet dictionary. They then used this collection of
opinion words to predict the sentiment orientation of electronic
product reviews at a sentence level. The dictionary-based approach
typically counts the numbers of positive and negative opinion
words in a sentence. If positive opinion words prevail, the orientation of the sentence is positive and otherwise negative. Sentiment
analysis can be conducted not only at a sentence level, but also
at others levels: document-, paragraph-, or attribute-level. As the
level of granularity increases, so does its complexity. Attributelevel sentiment analysis aims to associate opinions associated with
certain features (Chiu et al., 2015).
Sentiment analysis has been successfully applied in various
contexts, such as detecting inﬂuenza outbreak (Culotta, 2010),
determining overall trends in the level of happiness (Dodds et al.,
2011), predicting movie box ofﬁce revenues (Asur and Huberman,
2010), and understanding some consumer opinions on tourism and
hospitality related products (Chiu et al., 2015; Claster et al., 2013;
Duan et al., 2013). Chiu et al. (2015) developed a Chinese sentiment
analysis method to extract opinions toward various hotel attributes
in Chinese blogs. Duan et al. (2013) employed the sentiment analysis technique to decompose user reviews into ﬁve dimensions
of SERVQUAL (i.e. tangible, reliability, assurance, responsiveness,
and empathy), and measured consumers’ sentiments toward hotel
service quality. They also included the ﬁve dimensions to determine their effects on consumer’s satisfaction and review generating
behaviors. Their study results showed that consumers’ sentiments
on tangibles had the strongest impact on both dependent outcomes. In the tourism context, Claster et al. (2013) utilized the
technique to visualize the temporal sentiment ﬂuctuations associated with three tourist destinations: Thailand, Sri Lanka, and
Mexico. The researchers advocated sentiment analysis as a viable
tool to understand and benchmark tourist’s sentiments across
destinations.

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

The rising popularity of sentiment analysis in research can be
attributed to its unique advantages. First, as compared to manual coding, computer-aided sentiment analyses are not only more
efﬁcient, but also produce comparable results. Capriello et al.
(2013) compare the efﬁciency of manual content coding and two
computer-aided sentiment analyses techniques in analyzing 800
travel reviews of former farm stay guests. They found that all three
analyses produce similarly reliable results. Second, as compared
to traditional methods (e.g. surveys or focus groups), sentiment
analysis can effectively reduce cost, time, and manual labor by
using automatic algorithms to sort through text (Chiu et al., 2015).
Researchers contend that sentiment analysis, when used in analyzing a free ﬂow of user generated content, can discover dynamic and
unforeseen insights beyond the numerical rating of pre-determined
categories in surveys (Ghose and Ipeirotis, 2009). In addition, selfreport measurements, such as in satisfaction surveys, often suffer
from recall bias and question framing bias (Peterson and Wilson,
1992). Sentiment analysis of tweets can overcome these shortcomings, as Twitter users tend to post real time expressions that
reﬂect their in-the-moment experience.
Despite its advantages, sentiment analysis also has many drawbacks. As Pang and Lee (2008) pointed out, sentiment analysis is
domain and event dependent. Words considered positive in one
domain might not be so in another area. Sarcasm is an obvious
example of this limitation. Despite its inherent disadvantages, the
technique still appears as a promising tool for researchers and
industry practitioners. Wang et al. (2013) reviewed previous studies, and reported that the analysis technique yields a rather high
accuracy rate (roughly 70% to 80% accuracy rate in training-test data
matching tasks). The objective of sentiment analysis is to obtain
useful insight from a large quantity of aggregated data, rather than
perfect classiﬁcation of all data points.

total, 34 properties (and associated Twitter accounts) were used
as part of this study. The list of accounts/properties is provided in
Appendix A, along with their Twitter account parameters.
To compare the social media satisfaction scores, a sentiment
index was built using the data mined tweets. The sentiment score
was constructed by scoring the tweet text for positive and negative
words using Hu and Liu’s (2004) sentiment lexicon. The sentiment
score followed method used in Elder et al. (2012), where each positive word added a point to the total score and each negative word
removed a point–there are no points added or subtracted for neutral words. The properties were then compared on two criteria:
the average (mean) tweet sentiment score for any tweet that mentioned their corporate username (average score) and the ratio of
positively scored tweets to negatively scored tweets for any tweet
that mentioned their corporate username (ratio score). The average
score is deﬁned more formally as,
Average ScoreRt =

 n


PosRit −

n


i=1


NegRit

/nRt

i=1

where the average score at time = t is the difference between the
sum of positive (Pos) individual words appearing in tweet i of n total
tweets, and negative (Neg) individual words appearing in tweets
(also from i of n), divided by the total number of tweets, for any
tweet that mentions resort R.
The ratio score is deﬁned more formally as,

i=1
Ratio ScoreRt =  n

PTRit

i=1
NTRit
n

where,



PTRit =
3. Methodology
This study uses a dictionary-based method to analyze social
media microblogging data from Twitter. Data for this study was
acquired from the Twitter application programming interface (API).
The Twitter API is a backend server that warehouses all individuals’
tweets and enables data collection by the public. The site allows
roughly the past week of historical data to be extracted from its
database (Twitter, 2015). While there are many methods that can
be used to mine this data, we follow an approach adapted from Elder
et al. (2012), and use the TwitteR package for the R programming
language (Gentry, 2013). We also use the most recent version of
Hu and Liu’s (2004) sentiment lexicon to determine the sentiment
orientation of the online text.
TwitteR provides a well-documented and accessible means to
extract data into a commonly used data mining statistical program (KDnuggets, 2012). TwitteR is useful for many different data
mining procedures, but one form that can be particularly useful
to tourism ﬁrms is the searchTwitter function, which retrieves any
tweet containing a speciﬁed text string. This can, for example, be
used to identify any tweet mentioning a corporate Twitter account.
A searchTwitter query for “@hotelxyz” will return any tweet, by any
Twitter user, who mentioned that hotel’s username.
The data mining frame used in this study was to retrieve any
tweet that included an active integrated-resort username from the
Las Vegas metropolitan area. We deﬁne the scope of Las Vegas as
the Las Vegas Valley metropolitan area, and include only Twitter
veriﬁed accounts or non-veriﬁed accounts with at least 5000 followers. Las Vegas is a useful market for examining this data mining
procedure since it contains a large cluster of different branded hospitality ﬁrms that face similar market-wide conditions, and many
of these properties have a large number of Twitter followers. In


NTRit =

1 if PosRit − NegRit > 0
Otherwise

1 if PosRit -NegRit < 0

Otherwise

The ratio score is computed as the ratio of overall positive tweets
(PT) i of n at time t, to overall negative tweets (NT) i of n at time
t. More simply, in a given time period, the sum of all the positive
tweets for a given operator are compared to the sum of all negative tweets, and any neutral tweets are discarded. We illustrate
these two scoring methods with a short example of two hypothetical tweets: 1) “Had a great time @AriaLV. Liked the hotel
despite the noise at night,” and 2) “The @AriaLV buffet was awful.”
Both these examples would be linked to the @AriaLV corporate
account. The ﬁrst tweet had two positive words (underlined) and
one negative word (italicized), and the second tweet has one negative word. The average score for @AriaLV would therefore be:
((2 + 0) − (1 + 1)) /2 = 0, since there are two positive words in the
ﬁrst tweet and none in the second, along with one negative word
in each tweet. The ratio score would be: 1/1 =1, since there is one
overall positive tweet and one overall negative tweet.
The average score provides a useful baseline to understand
each ﬁrm’s relative sentiment, while the ratio score provides a
more meaningful overview of the most polarizing opinions, since it
ignores tweets with sentiment scores of zero. The internal validity
of the scores is compared using Cronbach’s alpha, and concurrent
validity is tested by comparing the measures to online rankings
from travel review site, TripAdvisor, another source of EWOM.
While we expect some levels of positive correlation between all
of these metrics, we also expect a reasonable amount of variation
that reﬂects different response time, demographics, and indirect
measurement involved in a Twitter based interaction.

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

Table 1
Property twitter sentiment scores.

Aria Las Vegas
Tropicana Las Vegas
Venetian Las Vegas
Bellagio Las Vegas*
Paris Las Vegas
Caesars Palace
Red Rock Las Vegas
M Resort Spa Casino
Luxor Hotel & Casino
Monte Carlo Resort
Wynn Las Vegas
Treasure Island
Circus Circus
Flamingo Las Vegas
Palazzo Las Vegas
LVH Hotel & Casino
South Point Hotel
Stratosphere Hotel
Harrah’s Las Vegas
The Mirage
The Cosmopolitan
Rio Las Vegas
Mandalay Bay Resort
NYNY Vegas
Palms Casino Resort
Golden Nugget
Riviera Las Vegas
The Quad Las Vegas
Excalibur Las Vegas
Hooters Casino Hotel
Planet Hollywood
MGM Grand Hotel
Hard Rock Hotel LV
Bally’s Las Vegas
*
**

Tweet
volume

Avg sentiment
score

Avg sentiment
rank

Pos/Neg ratio
score

Pos/Neg
ratio rank

TripAdvisor
relative rank

TripAdvisor
score (%)

0.69
0.69
0.68
0.66
0.64
0.63
0.61
0.59
0.58
0.57
0.56
0.55
0.55
0.55
0.53
0.52
0.52
0.52
0.51
0.51
0.48
0.46
0.45
0.44
0.42
0.42
0.41
0.41
0.41
0.40
0.36
0.35
0.34
0.30

10.72
11.10
8.66
9.60
9.90
6.08
10.22
5.96
6.10
6.81
6.79
6.66
5.68
5.42
7.04
5.86
5.88
6.05
4.03
4.90
5.20
4.61
5.78
5.08
5.24
3.13
5.23
2.57
5.53
4.64
3.39
4.41
2.74
2.62

N/A**

The Bellagio changed their corporate Twitter account on October 8, from @bellagiolv to @bellagio, which yielded unreliable data for period 2.
M Resort was not ranked relative to the other listed resorts by TripAdvisor.

A pilot test of the data extraction and cleaning procedure was
conducted on August 16, 2013, which led to some analysis modiﬁcations. Based on ﬁndings from the pilot test, tweets that were sent
from certain source platforms were pre-screened, since there was a
high percentage of fake accounts that could be identiﬁed on certain
outgoing platforms (e.g. constant account activity over several 24 h
periods). Twitter is similar to other online sites in terms of false
reviews, such as Yelp or TripAdvisor, but the richness in data may
yield a more effective means to screen some of these false accounts.
If some valid accounts were eliminated as a result of this prescreening, we expect this bias to be consistent across all properties.
We expect this outcome to be much less biased than the alternative, as the overly positive “shill” accounts all favored properties
belonging to a single corporate entity. We also chose to remove
retweeted messages—a single message that was retweeted from
many accounts could bias the results toward a single user’s opinion.
In particular, this bias would lean toward celebrities and corporate
accounts with large numbers of followers that can retweet messages. After examining the tweets’ scores in the pilot data, several
words were then added to the lexicon to reﬂect commonly used
positive/negative words that were absent from the lexicon. These
were primarily short-hand abbreviations of full words (e.g. “thx”) or
emoticons (e.g. “:-)”), which may be a reﬂection of the 140 character
limit of Twitter comments.
Collection dates then continued on August 23, August 30,
September 6, and September 13 in 2013 (referred to as period
one) as the Twitter Search API allows for roughly the past week
of data to be extracted at a given time. After analyzing the period
one data, we chose to add an additional set of data collection points
on October 25, November 1, November 8, and November 15 in 2013

(period two) for further tests of consistency. In total, 34,315 tweets
were mined during this procedure. There were 1083 tweets with
errors in decoding that were removed–this was primarily a result
of non-standard character use, such as accented letters in foreign
languages. A further 2765 duplicate tweets were removed, as the
multiple mining points of the Twitter API created the possibility
of redundancies. In total, 31,550 tweets were used in the ranking
analysis.
4. Results
The property sentiment scores over the eight weeks of data
collection are provided in Table 1. In general, positive sentiment
outweighs negative sentiment, as all ﬁrm had positive average sentiment scores, and positive ratio scores. On average, there were
928 tweets per account, with the MGM Grand account having the
most tweets (2832), and Hooter’s casino the fewest (1 8 7). While
many commonly known luxury brands such as Aria, Bellagio, and
Venetian appear in the top ﬁve average sentiment score rank, less
well known properties, such as Tropicana or the local market serving Red Rock and M Resort, also appear within the top ten. The
appearance of these ﬁrms suggests that these metrics are capturing
more information than simple views on hotel rankings. Indeed, Liu
(2012) describes sentiment as opinions, sentiments, evaluations,
appraisals, attitudes, and emotions, which is consistent with our
observations in analyzing resort tweets.
While the average score and the ratio score ranks had a strong
positive correlation with one another, Spearman’s  = 0.893, n = 34,
p < 0.001, a close examination of the scores reveals that several
ﬁrms had meaningfully different ranks, depending on the metric.

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

Given that average scores are positive, a higher number of tweets
that are neutral (sentiment score of zero) will reduce the average, but have no effect on the ratio score rank. This may explain
part of the difference in relative ranking. It is unclear if tweets that
score exceptionally high or low (i.e. multiple lexicon words) reﬂect
a stronger emotion, but this certainly could also be an explanation
of the difference in rankings. For example, properties that scored
relatively higher on the average score than the ratio score may
have a subset of customers with disproportionally strong feelings
of goodwill.
To examine the effectiveness of the sentiment scoring lexicon in
capturing week to week changes, we individually examined tweets
that appeared for properties with the highest and lowest average
scoring weeks of the study. The tweets in those given weeks were
reviewed by the authors for any consistent content themes that
may have face validity. To provide better context for the reader,
we describe the top three and bottom three average scoring weeks
that appear in the study for illustrative purposes–the date provided
reﬂects the data extraction date, and thus includes tweets up to
(roughly) seven days prior.
4.1. Positive events
The highest scoring extraction occurred on August 30 for Caesar’s Palace (192 tweets). This week captured a promotion by Matt
Goss (a Caesar’s Palace performer), that allowed the winner to
attend his performance and after-party on the property, which produced many positive responses from accounts where the user did
not even appear to be in Las Vegas. For example, “@mattgoss @CaesarsPalace Wow what an amazing prize.whoever wins will be so
lucky.goodluck everyone!”
The second highest scoring extraction occurred on November
8 for Red Rock Las Vegas (50 tweets). This casino, which generally serves local patrons rather than tourists, consistently scored
high during the study, but during this week it had many positive
tweets associated with promotion of an onsite event commemorating the 20th anniversary of the Ultimate Fighting Championship, for
example, “Celebrate 20 yrs of @UFC at @UltimatePoker Hall of Fame
Game w @ChuckLiddell @realroyce & more @redrockcasino 11/1 .
The third highest scoring extraction occurred on September 6 for
Harrah’s Las Vegas (43 tweets). The scoring this week was buoyed
by a series of high scoring tweets regarding, “Toby Keith’s I Love
this Bar and Grill”, which contained positive lexicon words, but
may not have been directly used to describe sentiment and instead
are simply a reﬂection of the restaurant name. This is an important example of limitations associated with sentiment analysis, and
the need to continue to reﬁne algorithmic approaches in applied
settings.
4.2. Negative events
The most negative extraction occurred on November 8 for
Planet Hollywood (36 tweets). During this period, there were several tweets remarking a looming feud between Britney Spears
(the recently announced Planet Hollywood headline act) and Paris
Hilton (whose has the same initials as Planet Hollywood). Unsurprisingly, given the subjects, the substantive meaning behind
these passionate observers remains somewhat undecipherable. For
example, one tweet said, “so will @ParisHilton so her support for
@britneyspears by doing a video or will the be jealousy over @phvegas by stealing intials at PH”.
The second most negative extraction occurred on October 25
for Bally’s Las Vegas (95 tweets), and provides good evidence of
Twitter data analysis’ ability to capture real time sentiment, with
an opportunity to qualitatively read tweets for more context. The
low score this week reﬂects many negative tweets associated with

a shooting that occurred at a nightclub located inside of Bally’s,
for example, “BREAKING: @LVMPD conﬁrms three men shot inside
@BallysVegas. One suspect in custody.”
Lastly, the third most negative extraction occurred on November
1 for The Quad (23 tweets). The Quad was among the lowest rated
properties overall, with the second lowest tweet volume. As such, a
moderate increase in negative reviews associated with poor service
and on-site construction appeared to have affected reviews this
week. This example demonstrates the importance of having a reasonably sized sample to gauge extreme events.
4.3. Validity tests
To assist in evaluating the content validity of the sentiment
metrics, we computed Spearman correlations of the properties’
rankings based on the average sentiment rank and the ratio
sentiment rank against TripAdvisor’s relative rankings of the hotelcasinos and TripAdvisor’s average scores of the hotel-casinos from
May 2013. The average sentiment ranks have a moderate positive
correlation with TripAdvisor ranks, Spearman’s  = 0.447, n = 33,
p = 0.009, and TripAdvisor scores, Spearman’s  = 0.432, n = 33,
p = 0.012. Similarly, the ratio sentiment ranks have moderate positive correlations with TripAdvisor ranks, Spearman’s  = 0.513,
n = 33, p = 0.002, and TripAdvisor scores, Spearman’s  = 0.501,
n = 33, p = 0.003.
The moderate, but statistically signiﬁcant levels of correlation
between the data mined sentiment metrics and the TripAdvisor
metrics appears to be in line with the expectation of a reasonable
amount of both convergent validity and discriminant validity. The
expected convergent and discriminant validities between the Twitter and Tripadvisor metrics lie in the overlapping targeted users
and generated content across these two sites. The content on TripAdvisor is primarily hotel experience reviews written by existing
customers. However, Twitter circulates not only hotel reviews but
also any news regarding current events and marketing campaign.
The targeted users are any consumers, existing and prospective,
who have an interest in the hotel brands. Therefore, we expect
some degree of correlation between these new metrics and online
rankings, since both are indicative of the customer experience and
sentiment (convergent validity). However, aspects that are unique
to the Twitter medium (e.g. the public’s reactions to current events
and marketing campaigns) lead to the expectation that there will
still be some differences in relative rankings (discriminant validity).
Indeed, customer sentiment is a related but somewhat different measure than customer satisfaction. Some tweets may reﬂect a
level of satisfaction from a consumed service, such as this positively
scored tweet to Aria’s Twitter account, “@AriaLV Absolutely loved
the rooms! Luxury!” While other tweets may be more accurately
described as an aspirational or perceptional comment, such as the
following positively scored tweet by a user seemingly outside of Las
Vegas, “@MGMGrand It is 40 degrees in the lovely city of Pittsburgh,
so YES, I wish I was in Vegas right about now!! #ilovemgmgrand”.
As compared to TripAdvisor scores, which appear to mainly reﬂect
customers’ hotel experience perceptions and amenities, Twitter
tends to capture a broader public opinion (including individuals
that have not visited the property) and their attitudes related to
wider topics (e.g. events or marketing campaigns indirectly related
to the properties).
4.4. Reliability tests
To better understand the reliability of the sentiment metrics,
Cronbach’s alpha was computed from week to week, and from
period one to period two (which each aggregated four weeks
of data). Cronbach’s alpha for the eight-item weekly average
sentiment score (˛ = 0.68), the two-item period sentiment score

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

(˛ = 0.59), the eight-item weekly ratio sentiment score (˛ = 0.53),
and the two-item ratio sentiment score (˛ = 0.50), were all below
the commonly used threshold of .70 (Hair et al., 1998)—though
the ﬁrst measure was above the .60 threshold used for exploratory
research (Robinson et al., 1991). However, unlike typical scales, we
are attempting to develop a metric that can capture dynamic movement of consumer sentiment in real time. We therefore ﬁnd the
Cronbach alpha scores encouraging, in that there is a certain degree
of consistency (we are not observing purely statistical noise), but
there remains a reasonable amount of periodical inﬂection. We also
note that Cronbach’s alpha scores can be inﬂated by including a
higher number of items for comparison, which we have attempted
to limit in these analyses.
More stable measures of sentiment (not shown in Table 1)
could
such
n include metrics
 as the net sum of sentiment scores
n
PosRit −
NegRit , where ˛ = 0.88 over eight weekly
i=1
i=1
scores and ˛ = 0.79 from
nperiod one to period two; the sum of positively scored tweets ( i=1 PTRit ), where ˛ = 0.88 over eight weekly
scores and ˛ = 0.80 from period
none to period two; and/or the sum
of negatively scored tweets ( i=1 NTRit ), where ˛ = 0.82 over eight
weekly scores and ˛ = 0.77 from period one to period two. These
more stable metrics may be useful by management in evaluating
performance longitudinally (e.g. long-term performance of a social
media staff or marketing campaigns), rather than as a comparative
measure relative to competitors, since the total volume of tweets
is an important component of their resulting impact.

5. Discussion
This study described the usefulness of sentiment analysis of
social media sites to hospitality operators. In particular, this study
demonstrated how Twitter data could be analyzed through a costeffective application to Las Vegas integrated-resort casino accounts.
The study showed that using a relatively straightforward scoring
algorithm with a publically available lexicon will yield a set of
sentiment metrics that have reasonable reliability and convergent
validity with external hotel rankings. As compared to Tripadvisor’s
ranking, the sentiment scores appear to be reasonably effective
at capturing real-time information: not only existing customers’
review on their hotel stay, but also the public’s buzz toward important short-term events and marketing campaigns.
Our study contributes to hospitality and tourism literature in
several ways. First, this study makes a methodological contribution by demonstrating a new and innovative analytical method
in understanding the public’s sentiments toward hospitality ﬁrms,
using a large volume of Twitter text data. As social media empowers consumers with more self-expression opportunities, hospitality
scholars have shown tremendous interests in consumers’ EWOM
behaviors. The existing literature mainly focuses on the motivations for and the effects of EWOM. Sparse is the work that attempts
to tackle the challenges of distilling business intelligence with large
quantities of online data.
While sentiment analysis has been touted as a promising tool to
uncover new knowledge from the massive volume of unstructured
data in many disciplines, it is rarely explored in the hospitality and
tourism context. Most recently, Xiang et al. (2015) applied text mining to analyze a large volume of consumer reviews and identiﬁed
the key dimensions of hotel guest experience that are associated
with their satisfaction ratings. While their study summarized what
hotel guests talk about their experience (e.g. room, staff, breakfast
etc.), it did not capture how the guests feel about their stay in their
reviews. Sentiment analysis goes beyond simply extracting useful
information from text to extracting opinions and feelings embedded in the text (Anbananthen and Elyasir, 2013). Using Las Vegas
integrated resorts as a case study, our study demonstrated that

resort operators can use Twitter sentiment analysis to construct
a set of valid and reasonably reliable metrics to capture the public’s
opinions. Sentiment analysis is not only restricted to Twitter data in
the resort context. It is also readily applicable to analyzing text data
from other online sources in other industries, such as restaurants
or tourism destinations.
From the hospitality operators’ perspective, sentiment analysis enables hospitality companies to track public viewpoints on a
large scale by generating graphical summarizations of opinions.
It is a tool that can be used to visualize a trajectory of the public buzz around a company by comparing changes in scores over
time and against other ﬁrms. By leveraging the temporal information associated with Twitter data, hospitality marketing managers
can effectively identify potential trends and publicity crises and
thus take timely actions. Sentiment analysis also helps marketing
managers to detect and exclude “ﬂames” (overly heated or antagonistic languages) in social communication, and provides support
to enhance antispam systems (Cambria et al., 2013). In this study,
the sentiment analysis revealed that some Las Vegas resorts are
clearly outperforming others in terms of engagement on social
media. Bally’s and Tropicana are both Las Vegas Strip properties
with similar amenities and TripAdvisor ratings, but Tropicana has
clearly outperformed Bally’s in terms of its presence on Twitter
and users’ willingness to engage with the brand through social
media.
Sentiment analysis can provide a quantiﬁable measure of
real-time marketing campaign impacts complementary to other
traditional research methods. Traditionally, the effectiveness of a
marketing campaign is evaluated through direct rating and brand
awareness tests using survey method or focus groups. While these
methods generate valid results, they can be costly, time consuming,
and require customers to rate categories that are pre-determined
by researchers or managers. Such a rating system might not necessarily capture the responses that are personally meaningful to
customers, and almost certainly will suffer from self-selection and
recall bias. While the effectiveness of marketing campaign can be
also gauged by sales effect (Kotler et al., 2012), this method is often
used in the post-campaign period, and has a time lag between the
initial campaign and resultant sales. Unlike these methods, sentiment analysis is able to capture the real-time and naturalistic
publicity impacts of current campaigns from a vast consumer base.
For example, Caesars’ Palace saw a surge of positive sentiment in
August 2013 right after a promotion to attend a performers and
after-party. Sentiment analysis has the advantage of immediacy
and cost efﬁciency over the traditional methods, and it is a complementary tool to justify the return on investment of marketing
campaigns.
There are many limitations to the use of sentiment analysis.
We note that the Twitter API provides a limited search capacity
regarding returned data volume and time frame. The tweets which
are in private setting are not retrievable through the API. Analysis is limited to the extent that social media platform owners are
willing to share their data, or that data can be legally mined. Our
study focused on a speciﬁc category of hospitality ﬁrms (integrated
casino resorts), within a single market (Las Vegas, NV). Application of this method in other industries or locations may lead to
different results, necessitating methodological modiﬁcations. For
example, small hotel properties may not have sufﬁcient Twitter
volume to conduct such analyses, and sentiment analysis may
therefore be limited to large corporate accounts with well established brands.
While the lexicon used in this study may have measurement
validity across a large sample of tweets, smaller samples or individual comments can be easily miscategorized. Future studies should
focus on improving the content validity of these text mining and
classiﬁcation algorithms, and establishing rules of thumb to guide

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

managers toward meaningful sample sizes. Where account mentions are infrequent, a more qualitative/ethnographic approach
to examining tweets may yield a richer understanding of consumer attitudes and emotions. We caution that social media, and
Twitter in particular, may not be representative of a ﬁrm’s overall target market. Twitter users skew toward younger and urban
groups (Vaynerchuk, 2013), therefore managers should compare

demographics captured in Twitter mining to those that are representative of their typical or target consumer. Indeed, Twitter
sentiment analysis certainly does not replace traditional market
research tools like guest satisfaction surveys, but it can be a useful
complementary tool.
Appendix A. Twitter account information

Twitter username

Twitter name

Followers

Following

Tweets

Veriﬁed
account

@AriaLV
@BallysVegas
@BellagioLV*
@CaesarsPalace
@CircusVegas
@Cosmopolitan LV
@ExcaliburVegas
@FlamingoVegas
@GoldenNuggetLV
@hardrockhotellv
@HarrahsVegas
@HootersCasinoLV
@LuxorLV
@LVHHotelCasino
@LVStratosphere
@MandalayBay
@MGMGrand
@MonteCarloVegas
@MResort
@NYNYVegas
@PalazzoVegas
@Palms
@ParisVegas
@phvegas
@quadvegas
@redrockcasino
@RioVegas
@RivieraLasVegas
@southpointlv
@TheMirageLV
@TIvegas
@TropLV
@VenetianVegas
@WynnLasVegas

Aria Las Vegas
Bally’s Las Vegas
Bellagio Las Vegas
Caesars Palace
Circus Circus
The Cosmopolitan
Excalibur Las Vegas
Flamingo Las Vegas
Golden Nugget
Hard Rock Hotel LV
Harrah’s Las Vegas
Hooters Casino Hotel
Luxor Hotel & Casino
LVH Hotel & Casino
Stratosphere Hotel
Mandalay Bay Resort
MGM Grand Hotel
Monte Carlo Resort
M Resort Spa Casino
NYNY Vegas
Palazzo Las Vegas
Palms Casino Resort
Paris Las Vegas
Planet Hollywood
The Quad Las Vegas
Red Rock Las Vegas
Rio Las Vegas
Riviera Las Vegas
South Point Hotel
The Mirage
Treasure Island
Tropicana Las Vegas
Venetian Las Vegas
Wynn Las Vegas

96,519
12,478
58,884
82,496
17,151
137,164
54,345
26,456
11,366
63,108
18,234
15,079
91,911
29,741
9,300
78,788
101,395
64,229
18,535
43,234
32,107
76,362
23,944
54,983
10,064
17,088
33,388
12,221
15,248
43,226
26,682
21,907
42,111
488,061

46,389
74,429
14,924
13,921
22,503
28,517
25,761
12,674
10,196
25,147

11,304
34,039
15,475
11,045
12,059
12,301
24,229
10,900
15,058

Yes
No
Yes
Yes
Yes
Yes
Yes
No
No
No
No
No
Yes
No
No
Yes
Yes
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
Yes
No
No
Yes

Note: Figures were recorded at the beginning of the study.
*
The Bellagio changed their corporate Twitter account on October 8, from @bellagiolv to @bellagio, which yielded limited data for period 2.

K. Philander, Y. Zhong / International Journal of Hospitality Management 55 (2016) 16–24

References
Akehurst, G., 2009. User generated content: the use of blogs for tourism
organizations and tourism consumers. Serv. Bus. 3 (1), 51–61.
Anbananthen, K.S.M., Elyasir, A.M.H., 2013. Evolution of opinion mining. Aust. J.
Basic Appl. Sci. 7 (6), 359–370.
Asur, S., Huberman, B.A., 2010. Predicting the future with social media. In: Web
Intelligence and Intelligent Agent Technology (WI-IAT), 2010 IEEE/WIC/ACM
International Conference on, August 2010 vol. 1, pp. 492–499, http://dx.doi.
org/10.1109/WI-IAT.2010.63, Retrieved from IEEE Xplore.
Boyd, D., Crawford, K., 2012. Critical questions for big data: provocations for a
cultural, technological, and scholarly phenomenon. Inf. Commun. Soc. 15 (5),
662–679.
Browning, V., So, K.K.F., Sparks, B., 2013. The inﬂuence of online reviews on
consumers’ attributions of service quality and control for service standards in
hotels. J. Travel Tourism Mark. 30 (1–2), 23–40.
Cambria, E., Schuller, B., Xia, Y., Havasi, C., 2013. New avenues in opinion mining
and sentiment analysis. IEEE Intell. Syst. 28 (2), 15–21.
Capriello, A., Mason, P.R., Davis, B., Crotts, J.C., 2011. Farm tourism experiences in
travel reviews: a cross-comparison of three alternative methods for data
analysis. J. Bus. Res. 66, 778–785.
Cha, M., Haddadi, H., Benevenuto, F., Gummadi, P.K., 2010. Measuring user
inﬂuence in twitter: the million follower fallacy. In: The Conference
Proceeding of the Fourth International AAAI Conference on Weblogs and Social
Media, Washington, DC, pp. 10–17.
Chan, N.L., Guillet, B.D., 2011. Investigation of social media marketing: how does
the hotel industry in Hong Kong perform in marketing on social media
websites? J. Travel Tourism Mark. 28 (4), 345–368.
Chen, H., Chiang, R.H., Storey, V.C., 2012. Business intelligence and analytics: from
big data to big impact. MIS Q. 36 (4), 1165–1188.
Chiu, C., Chiu, N.H., Sung, R.J., Hsieh, P.Y., 2015. Opinion mining of hotel
customer-generated contents in Chinese weblogs. Curr. Issues Tourism 18 (5),
477–495.
Claster, W., Pardo, P., Cooper, M., Tajeddini, K., 2013. Tourism, travel and tweets:
algorithmic text analysis methodologies in tourism. Middle East J. Manage. 1
(1), 81–99.
Culotta, A., 2010. Detecting Inﬂuenza Outbreaks by Analyzing Twitter Messages,
Retrieved from Cornell University Library http://arxiv.org/abs/1007.4748.
Ding, X., Liu, B., Yu, P.S., 2008. A holistic lexicon-based approach to opinion mining.
In: Proceedings of the 2008 International Conference on Web Search and Data
Mining, ACM, pp. 231–240.
Dodds, P.S., Harris, K.D., Kloumann, I.M., Bliss, C.A., Danforth, C.M., 2011. Temporal
patterns of happiness and information in a global social network:
Hedonometrics and Twitter. PLoS ONE 6 (12), 1–26.
Duan, W., Cao, Q., Yu, Y., Levy, S., 2013. Mining online user-generated content:
using sentiment analysis technique to study hotel service quality. In: System
Sciences (HICSS), 2013 46th Hawaii International Conference on, IEEE, January
2013, pp. 3119–3128.
Dugann, M., Smith, A., December 2013. Social Media Update 2013. Pew Research
Center, Retrieved from
http://pewinternet.org/∼/media//Files/Reports/2013/Social%20Networking%
202013 PDF.pdf.
Elder, J., Hill, T., Delen, D., Fast, A., 2012. Practical Text Mining and Statistical
Analysis for Non-structured Text Data Applications. Academic Press, Waltham,
MA, pp. 133–142.
Gentry, J., 2013. Package ‘twitteR’. R Core Development Team, Retrieved from
http://cran.r-project.org/web/packages/twitteR/twitteR.pdf..
Ghose, A., Ipeirotis, P., 2009. The EconoMining project at NYU: studying the
economic value of user-generated content on the internet. J. Revenue Pricing
Manage. 8 (2), 241–246.
Gibbs, G., 2007. Analyzing Qualitative Data. Sage, Thousand Oaks, London.
Grandoni, D., Dec 2013. One Statistic That Shows How Small Twitter Really is. The
Hufﬁngton Post, Retrieved from
http://www.hufﬁngtonpost.com/2013/12/19/twitterstatistics n 4469054.html.
Hair, Joseph F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data
Analysis. Prentice Hall, Upper Saddle River, NJ, pp. 577–664.
Hays, S., Page, S.J., Buhalis, D., 2013. Social media as a destination marketing tool:
its use by national tourism organisations. Curr. Issues Tourism 16 (3), 211–239.
Hennig-Thurau, T., Gwinner, K.P., Walsh, G., Gremler, D.D., 2004. Electronic
word-of-mouth via consumer-opinion platforms: what motivates consumers
to articulate themselves on the internet? J. Interact. Mark. 18 (1), 38–52.
Hennig-Thurau, T., Malthouse, E.C., Friege, C., Gensler, S., Lobschat, L.,
Rangaswamy, A., Skiera, B., 2010. The impact of new media on customer
relationships. J. Serv. Res. 13 (3), 311–330.
Hennig-Thurau, T., Hofacker, C.F., Bloching, B., 2013. Marketing the pinball way:
understanding how social media change the generation of value for consumers
and companies. J. Interact. Mark. 27, 237–241.
Hu, M., Liu, B., 2004. Mining and summarizing customer reviews. In: Proceedings
of the Tenth ACM SIGKDD International Conference on Knowledge Discovery
and Data Mining, ACM, August 2004, pp. 168–177, http://dx.doi.org/10.1145/
1014052.1014073.

View publication stats

KDnuggets, 2012. Poll Results: Top Analytics, Data Mining, Big Data Software Used,
Retrieved from KDnuggets http://www.kdnuggets.com/2012/05/topanalytics-data-mining-big-data-software.html..
Kim, J., Connolly, D., 2013. The 2013 customer engagement technology study.
Hospitality Technol., 3–22, Retrieved from
http://hospitalitytechnology.edgl.com/reports/143-Findings-on-How-YourCompetitors-are-Engaging-Guests–2013-Customer-Engagement-TechStudy87653.
Kirilenko, A.P., Stepchenkova, S.O., 2014. Public microblogging on climate change:
one year of twitter worldwide. Global Environ. Change 26, 171–182.
Kotler, P., Bowen, J., Makens, J., 2012. Promotion products: communication and
promotion policy and advertising. In: Marketing for Hospitality and Tourism,
sixth ed. Prentice Hall, Upper Saddle River, NJ, pp. 361–395.
Krikorian, Rafﬁ, 2013. New tweets per second record, and how! Twitter Eng. Blog,
Retrieved from
https://blog.twitter.com/2013/new-tweets-per-second-record-and-how.
Kwak, H., Lee, C., Park, H., Moon, S., 2010. What is Twitter, a social network or a
news media? In: Proceedings of the 19th International Conference on World
Wide Web, ACM, April 2010, pp. 591–600, http://dx.doi.org/10.1145/1772690.
1772751.
Ladhari, R., Michaud, M., 2015. eWOM effects on hotel booking intentions,
attitudes, trust, and website perceptions. Int. J. Hospitality Manage. 46, 36–45.
Leung, D., Law, R., van Hoof, H., Buhalis, D., 2013. Social media in tourism and
hospitality: a literature review. J. Travel Tourism Mark. 30 (1–2), 3–22.
Litvin, S.W., Goldsmith, R.E., Pan, B., 2008. Electronic word-of-mouth in hospitality
and tourism management. Tourism Manage. 29 (3), 458–468.
Liu, B., 2012. Sentiment analysis and opinion mining. Synth. Lect. Hum. Lang.
Technol. 5 (1), 1–167.
Mauri, A.G., Minazzi, R., 2013. Web reviews inﬂuence on expectations and
purchasing intentions of hotel potential customers. Int. J. Hospitality Manage.
34, 99–107.
Nielsen, December 2012. State of the Media: The Social Media Report 2012,
Retrieved from http://www.nielsen.com/us/en/reports/2012/state-of-themedia-the-social-media-report-2012.html.
Pang, B., Lee, L., 2008. Opinion mining and sentiment analysis. Found. Trends Inf.
Retrieval 2 (1–2), 1–135.
Peterson, R.A., Wilson, W.R., 1992. Measuring customer satisfaction: fact and
artifact. J. Acad. Mark. Sci. 20 (1), 61–71.
Robinson, J.P., Shaver, P.R., Wrightsman, L.S., 1991. Criteria for scale selection and
evaluation. Meas. Pers. Soc. Psychol. Attitudes 1, 1–16.
Rylander, R.G., Propst, D.B., McMurtry, T.R., 1995. Nonresponse and recall biases in
a survey of traveler spending. J. Travel Res. 33 (4), 39–45.
Schuckert, M., Liu, X., Law, R., 2015. Hospitality and tourism online reviews: recent
trends and future directions. J. Travel Tourism Mark. 32 (5), 608–621.
Sparks, B.A., Browning, V., 2011. The impact of online reviews on hotel booking
intentions and perception of trust. Tourism Manage. 32 (6), 1310–1323.
Sparks, B.A., Perkins, H.E., Buckley, R., 2013. Online travel reviews as persuasive
communication: the effects of content type, source, and certiﬁcation logos on
consumer behavior. Tourism Manage. 39, 1–9.
Thelwall, M., Buckley, K., Paltoglou, G., 2011. Sentiment in Twitter events. J. Am.
Soc. Inf. Sci. Technol. 62 (2), 406–418.
Turney, P.D., Littman, M.L., 2003. Measuring praise and criticism: inference of
semantic orientation from association. ACM Trans. Inf. Syst. (TOIS) 21 (4),
315–346.
Twitter, 2015. FAQ: Twitter developers, Available at: https://dev.twitter.com/faq.
Vaynerchuk, G., 2013. Jab, Jab, Jab, Right Hook: How to Tell Your Story in a Noisy
Social World. Harper Collins, New York, NY.
Vermeulen, I.E., Seegers, D., 2009. Tried and tested: the impact of online hotel
reviews on consumer consideration. Tourism Manage. 30 (1),
123–127.
Wang, J., Gu, Q., Wang, G., 2013. Potential power and problems in sentiment
mining of social media. Int. J. Strateg. Decis. Sci. (IJSDS) 4 (2), 16–26.
Weil, K., 2010. Measuring tweets. In: Twitter Blog, Retrieved from
https://blog.twitter.com/2010/measuring-tweets.
Weng, J., Lim, E.P., Jiang, J., He, Q., 2010. Twitterrank: ﬁnding topic-sensitive
inﬂuential twitterers. In: Proceedings of the Third ACM International
Conference on Web Search and Data Mining, ACM, February 2010, pp. 261–270.
Xiang, Z., Schwartz, Z., Gerdes, J.H., Uysal, M., 2015. What can big data and text
analytics tell us about hotel guest experience and satisfaction? Int. J.
Hospitality Manage. 44, 120–130.
Xie, H., Miao, L., Kuo, P.J., Lee, B.Y., 2011. Consumers’ responses to ambivalent
online hotel reviews: the role of perceived source credibility and
pre-decisional disposition. Int. J. Hospitality Manage. 30 (1), 178–183.
Ye, Q., Law, R., Gu, B., 2009. The impact of online user reviews on hotel room sales.
Int. J. Hospitality Manage. 28 (1), 180–182.
Ye, Q., Law, R., Gu, B., Chen, W., 2011. The inﬂuence of user-generated content on
traveler behavior: an empirical investigation on the effects of
e-word-of-mouth to hotel online bookings. Comput. Hum. Behav. 27 (2),
634–639.
Zhang, Z., Ye, Q., Law, R., Li, Y., 2010. The impact of e-word-of-mouth on the online
popularity of restaurants: a comparison of consumer reviews and editor
reviews. Int. J. Hospitality Manage. 29 (4), 694–700.