ArticlePDF Available

Abstract and Figures

Tourist destinations are increasingly affected by the travel-related information shared through the Web. More and more people first check the previous experiences of other customers before doing their own decision-making. This paper explores the image of travel destinations by analysing the content of opinions shared using sentiment analysis techniques. A sentiment score is obtained and analysed considering several tourist features of the travel destination as well as the usefulness of shared opinions. A well known e-word of mouth community and the city of Barcelona have been used as a case study. The results obtained reveal the sentiment orientation towards the city of Barcelona and its tourist offer. Finally, a sensitivity analysis related to the calculation of the sentiment score is included.
Content may be subject to copyright.
Monitoring travel-related information on Social
Media through sentiment analysis
M.R. Gonzalez-Rodriguez, M. R. Martinez-Torres,
Business and Management department
University of Seville
Seville, Spain,
S. L. Toral
E. S. Ingenieros
University of Seville
Seville, Spain
Abstract—Tourist destinations are increasingly affected by
the travel-related information shared through the Web. More
and more people first check the previous experiences of other
customers before doing their own decision-making. This paper
explores the image of travel destinations by analysing the content
of opinions shared using sentiment analysis techniques. A
sentiment score is obtained and analysed considering several
tourist features of the travel destination as well as the usefulness
of shared opinions. A well known e-word of mouth community
and the city of Barcelona have been used as a case study. The
results obtained reveal the sentiment orientation towards the city
of Barcelona and its tourist offer. Finally, a sensitivity analysis
related to the calculation of the sentiment score is included.
Keywords—Social Media; sentiment analysis; travel-related
information; tourism 2.0
I. I
Destination image is one of the topics that are widely
studied in the literature. This is because it is considered a
powerful tool for destination marketers in order to obtain a
high competitive advantage in the tourist markets [1], [2]. It is
commonly accepted that destination image highly influences
the destination choice as well as the future behavioural
purchase intentions [3].
Recently, studies on tourism have examined the image
formation process of the pre-visit experience using different
sources of information, generally beyond the destination
marketer organisations (DMOs) [4]. The image formation
prior to the visit has been highly influenced by the information
provided by travellers’ worth of mouth (WOM). This is
related not only to the service quality but also to the quality of
the experience in the tourism destination [5]. Thus, WOM has
been sought as a credible source of information for the
destination choice. On the other hand, service quality [6], [7]
and, more recently, the quality of experience [8] and perceived
value [9], [10] appear to be the key drivers for the intention to
revisit and the willingness to recommend a destination. This
leads to new destination image perceptions after the visit.
The development and acceptance of information
technologies have changed the sources of information for the
image formation process. This has been traditionally focused
on WOM, media, travel guidebooks, brochures and magazines
[11]. The growing trend of online communities [12] has
encouraged customers to interact and share experiences
through an increasing number of reviews, based not only on
product quality but also on service experiences [13]. As a
consequence, highly-involved virtual customers seek and learn
from online consumer reviews displayed in online travel
communities. This way, they can acquire information and
create their own travel experience for a tourist destination
[14]. Decision-making based on others' previous experience
can also lead to wrong decisions, as decision criteria can be
different for different customers. However, the information
quality in virtual communities is assured through peer
evaluation in terms of the utility score given by other
community members. Trust and previous experience are also
key mechanisms to reduce the uncertainty from users’ reviews
[15], [16]. Although the user posting the reviews can also
score the product or service under review, this study proposes
going further in the analysis of the content of shared reviews.
More specifically, this study proposes to measure a sentiment
score associated with each review and decide about its
sentiment (positive or negative) orientation. The sentiment
orientation can then be used to measure the image destination
in the case of travel experiences. It can also be correlated with
some other available information, such as scores received or
previous experience. The rest of the paper is organised as
follows: section II includes the most relevant work in this
field. Section III introduces the case study as well as the
methodology for calculating the sentiment score. Section IV
shows the results obtained and, finally, conclusions are
detailed in section V.
It is widely accepted that image is the ability to influence
customers’ perception of the products and services offered
[17] and therefore their purchase decision-making. However,
there seems to be a general agreement that the destination
image formation process is a complex task with no consensus
concerning the measurement process. In the service literature,
researchers admit that image can be seen as a
multidimensional construct formed by the cognitive and the
affective dimensions [18]. The cognitive component is
determined by the evaluation of the destination’s objective
attributes according to external stimulus, including sources of
information, past experience and WOM. These stimuli allow
2014 IEEE/ACM 7th International Conference on Utility and Cloud Computing
978-1-4799-7881-6/14 $31.00 © 2014 IEEE
travellers to gather knowledge or beliefs about a certain
destination [19]. The affective counterpart is focused on
feelings. These in turn depend on the socio-psychological
travel motivation and psychological and personal
circumstances [20]. The study of the cognitive and the
affective dimensions enable a better comprehension about the
mental representation of the idea of a certain destination.
The projected representation considered in the literature as
a “push factor” in the election of a tourist destination appears
to be of great interest for its adequate promotion and
commercialisation [21]. A positive mental representation
becomes a high-perceived costumer value influencing future
buying behaviour [22].
The destination image literature review reveals that most of
the studies have been traditionally focused on the image’s
cognitive component to formulate the idea in the tourists'
minds [23]. These studies emphasise the use of tangible
attributes related to the tourist resources of the specific
destination [24]. However, a growing number of studies also
highlight the importance of jointly evaluating the cognitive
and affective perception of a tourist destination [25], [26].
Moreover, Kim and Richardson [20] admit that in a tourist
context the affective attributes of a destination are even more
important than the evaluation of the tangible attributes.
Traditionally, two methodologies have been applied to
measure the destination image: structured and unstructured
techniques. Structured methodology is focused on the
assessment of a battery of relevant attributes for a destination
previously identified in the study by the researcher. Structured
methodology using Likert or semantic scales has been
frequently applied in the literature, including tangible
attributes [27] and affective attributes [28], [20] to measure
the destination image. However, it is widely recognised that
this technique cannot capture those attributes that could be
considered of special relevance by the respondents [29] and
the unique component of the destination [30]. On the other
hand, the unstructured technique based on open-ended
questionnaires apprehends the holistic components and unique
features of a destination image [31], [32]. But in this case, the
great variability in the individual description and impressions
of a destination makes any comparative analysis difficult.
Thus, as identified by Echtner and Ritichie [33], due to its
inherent mental representation complexity the measurement of
the destination image justifies using a combination of
structured and unstructured methodologies.
While past qualitative and quantitative research was
focused on defining the multi-attribute destination image,
recent studies have explored the role of information sources in
the creation of the destination brand image. Stokes and Lomax
[5] recognise the strong influence of WOM information in the
brand image creation. They suggest that DMOs and
entrepreneurs should have a high understanding of that
information source as part of their marketing strategy to create
an engaging, memorable and newsworthy travel experience.
Further evidence of the power of WOM as a brand image
builder can be also found in Hanlan and Kelly [34]. They
identified the dominant attributes image in the individual
mental representation. The rapid and growing development of
the Web2.0 providing efficient tools to create and share ideas
(eWOM) through the World Wide Web has led to a new
image measurement paradigm. Online travel communities,
blogs and forums encourage people to share information about
a wide variety of destinations. This helps other information
seekers to create their own destination image. Since the
information on the web is unstructured, a new field within
content analysis based on sentiment orientation is starting to
appear in the tourist literature [35], [36]. Marketer planners are
increasingly using sentiment analysis to develop their
marketing strategies according to the consumer attitudes
manifested through e-WOM communities. Until recently,
most research based on sentiment analysis was focused on
product reviews rather than on service reviews. This research
means to fill this gap in the literature by examining electronic
reviews in the online community Ciao as an information
source for the creation of a destination brand image.
The case study is based on a well-known e-WOM
community, covering a wide variety of products and services.
The next subsections introduces e-WOM communities as well
as the data collection and the methodology applied.
A. e-WOM communities
The emergence of Internet and the Web 2.0 has changed
the way in which users look for information and their purchase
decisions. Previously, much of this information was received
directly from companies or from friends or relatives in
informal conversations. However, traditional WOM has now
its digital counterpart with e-WOM. e-WOMs allow
consumers to share their experiences, exchange product- and
service-related information and socially interact with other
consumers [37]. e-WOM takes place in a computer-mediated
context and, in contrast with traditional WOM, conversations
are visible to the rest of the consumers. This fact makes these
communities ideal for researchers because plenty of
information can be obtained via the web. Typically, reviews or
posted experiences are rated by the rest of the community in
terms of their usefulness [38]. This information is useful for
consumers to distinguish malicious opinions. As Internet is a
relatively anonymous medium, vendors can be tempted to
manipulate opinions by over scoring their products or services
and persuading against competitors. Additionally, some data
about users who share reviews are also displayed, such as
previous experience posting information, activities, reputation
or the date they joined the community [39].
B. Data collection
Data were collected from the website This is a
well-known e-WOM community that covers a wide variety of
products and services. The website is organised in 28 main
categories. One of them is the category 'Travel', which in turn
is divided into continents, countries and cities within each
continent. This study is specifically focused on the city of
Barcelona, which is one of the top touristic destinations in
Spain. Basically, Ciao distinguishes four areas related to the
experience of travelling to Barcelona: Hotels, Attractions,
Restaurants and Pubs, Bars & Nightlife. Registered Ciao users
can freely post reviews about any of these subcategories
related to travelling to Barcelona. These reviews can receive a
score from the rest of the community and this is also publicly
displayed. Finally, some statistics about users are also
available, such as the date they joined the community, the
number of previous submitted reviews or a trust score given
by the number of community members who specifically trust a
given user. All this information as well as the content of the
reviews was collected using a web scraper developed in R.
The function readLines() from the base package, that reads
data from a URL, was used to access the shared reviews.
However, webpages are formatted in HTML code, and
accessed data contains both the webpage content and the
HTML tags. Therefore, it is necessary to parse the HTML file
using the htmlParse() function . This generates an R structure
representing the HTML tree. Once online webpages are
available as an R structure, meaningful data can be easily
identified using regular expressions that are also supported in
R, for instance, in packages such as XML. Table 1 summarises
the information collected.
Variable Description
text Body of the posted review
size Size of reviews in words
user Alias and link to the user statistics
Date Date on which the user became a member
of Ciao
Reviews Number of previously posted reviews
Trust Members trusting this user
Subcategory Hotels, Attractions, Restaurants,
Table 1. Information extracted from reviews posted in the
category Travel -> Spain -> Barcelona
C. Sentiment analysis
Sentiment analysis refers to detect and classify the
sentiments expressed by an opinion holder. There are two
main approaches to the problem of automatically extracting
sentiment. A lexical-based approach involves calculating the
semantic orientation for a document from the semantic
orientation of words or phrases in the document. This is based
on a predefined list of words, where each word is associated
with a specific sentiment. Machine learning techniques
involve building classifiers from labelled instances of texts or
sentences through a supervised classification process [40]. The
advantage of machine learning techniques is that they can
create trained models for specific contexts. Nevertheless, they
also require the availability of labelled data. This compromises
their applicability to new data. This study follows the first
approach based on AFINN-111, which is a list of English
words rated for valence with an integer between minus five
(negative) and plus five (positive) for 2477 word forms [41].
Although the original ratings vary between -5 and +5, a
reclassification in a lower number of categories was done to
clearly highlight the semantic orientation. As a result, four
categories were distinguished: very negative (Vneg, rating -5
and -4), negative (Neg, rating -3, -2, or -1), positive (Pos,
rating 1, 2, or 3) and very positive (Vpos, rating 4 or 5). For
each review, the number of words belonging to each category
was calculated, and the final sentiment score was obtained as:
VposwPoswNegwVnegwSentiment +++= 2112 )()(
w1 and w2 being two weight factors that consider the relative
importance of very positive or negative terms with respect to
the positive or negative terms. Obviously, it is required that
w2> w
1 to emphasise the weight of very positive or negative
terms. In general, people avoid being too extreme in their
opinions, so the presence of very positive or negative words
should be considered as a clear semantic orientation.
A total of 200 posted reviews about the city of Barcelona
posted were collected at using the web scraper
Attractions Hotels Pubs_Bar s_Nightlife R estaurants
Travel subcategories
Figure 1. Opinion distribution by travel subcategories.
The distribution of the opinions by travel subcategory is
also important to understand those attributes of the destination
that should be improved. Figure 1 shows how many reviews
each travel subcategory has received, revealing those tangible
aspects of the city which are more assessed by travellers.
Hotels, followed by Attractions and Restaurants are the
categories receiving more attention from travellers. In
contrast, Pubs, Bars and Nightlife received a lower number of
reviews. This fact can be explained by the wide variety of
pubs and bars in Barcelona which makes it more difficult for
users to become engaged in discussions about them.
The sentiment analysis approach proposed has been
applied to the reviews collected. As a result, and using the
referenced dictionary, 6660 terms belonging to the four
categories considered (VNeg, Neg, Pos, Vpos) were identified.
Table 2 and Figure 2 describe the distribution of the sentiment
reviews by travel subcategory.
VNeg Neg Pos VPos
Hotels 4 749 1887 89
Attractions 2 500 1360 104
Restaurants 4 582 1086 37
Pubs&clubs 0 77 171 8
Table 2. Distribution of sentiment reviews by travel
It can be noted that users don not tend to use very positive
or very negative terms. A positive bias towards the city of
Barcelona can also be observed, although the negative
sentiment scores deserve special attention by destination
marketing planners. In general, travellers appear to be
reluctant to give very polar opinions to any subcategory.
VNeg Neg Pos VPos
Rating scale of words
Restaura nts
Figure 2. Distribution of sentiment reviews by travel
Sentiment scores were calculated for the 200 reviews
about the city of Barcelona, considering the score described in
section III with weight factors w1=2 and w2=5.
-50 0 50 100 150 200
Sentiment score
Figure 3. Sentiment Scores histogram with fitted density
Figure 3 shows the histogram of sentiment scores with its
density curve superimposed. The result obtained exhibits a
clear right skewed distribution. This means that most travellers
have a positive orientation towards Barcelona as a tourist
destination. Additionally, and due to the chosen weights, the
sharp shape of the histogram means that most opinions are
concentrated in a narrow range of sentiment scores.
Several subsequent analyses have been performed to detect
if the sentiment score is affected by travel subcategories or the
perceived utility of reviews. Table 3 provides the analysis of
the variance (ANOVA) results to test for any significant
differences among the mean scores for the four travel
N Mean F value Sig
Hotels 93 29.04 1.084 0.357
Attractions 62 35.97
Restaurants 36 35.28
Pubs & clubs 9 25.33
Levene Test=1.4754, Sig. = 0.2225
Table 3. ANOVA results for subcategories of the city of
Levene's test is not significant. This means that the
homogeneity of variance assumption was not violated. The
ANOVA results show that there are no differences in mean
sentiment scores among the travel subcategories
The utility of reviews is another important aspect for
customers when making decisions about travelling. The
ANOVA test has been applied to detect possible differences in
sentiment mean scores among different utility perceptions.
Table 4 shows a significant F value, which indicates
differences in the sentiment scores’ means but it does not
provide a multiple pairwise comparison.
N Mean F value Sig
Helpful 42 16.76
Very Helpful 96 28.28
Exceptional 62 46.98
Levene Test=7.3868, Sig. = 0.0001
Table 4. ANOVA results for utility of reviews.
For this purpose, the Tukey post-hoc test has been used to
determine the differences in sentiment scores between each
pair of utility categories. The sentiment score mean is higher
whenever the review’s utility perceived by the online visitors
is also higher.
diff Sig.
helpful-exceptional -30.22 0.000
very helpful-exceptional -18.70 0.000
very helpful-helpful 11.52 0.023
Table 5. Results of the Tukey post hoc test.
This result indicates that users tend to evaluate positive
reviews better than negative reviews. This actually seems
logical, as positive reviews are helping users to find
accommodation or attractions, while negative reviews just
make them continue looking for other alternatives.
Several other continuous variables collected were finally
correlated with sentiment scores in Table 6, including (1)
Sentiment scores, (2) Number of previous reviews, (3) Trust,
(4) Size of reviews in words, (5) Time reviewers have been
members of Ciao.
(1) (2) (3) (4) (5)
(1) 1
(2) 0.18
(3) 0.14 0.75
(4) 0.57
(5) -0.03 -0.03 0.16 -0.06 1
Note: ***p<0.001; **p <0.01; *p<0.05
Table 6. Correlation Analysis
The sentiment score is positively correlated with the length
of reviews and the previous experience of the reviewers. The
first result seems to be logical, since the higher the length of
the text, the clearer its semantic orientation. The positive,
although weak, correlation with the number of previous
reviews means that experienced users tend to review those
services where they had a positive experience. Other results
show that experienced users tend to be trusted and they also
write more detailed reviews.
A sensitivity analysis was next performed to test the
influence of weights in the sentiment score calculation.
Several combinations of weights were chosen and applied,
considering the restriction of w2> w1. The histograms obtained
are shown in Figure 4.
w1=1, w2= 5 w1=1, w2= 10
w1=2, w2= 5 w1=2, w2= 10
w1=3, w2= 5 w1=3, w2= 10
w1=4, w2= 5 w1=4, w2= 10
0 100 200 300 400 0 100 200 300 400
Sentiment score
Figure 4. Sentiment Scores histogram with fitted density curve considering several weights combinations.
It can be noticed that, as weights become higher and the
difference between w2 and w1 is also higher, the density curves
exhibit a higher positive skew. Additionally, distributions
become flatter, with a greater variety of scores. In general, a
wider variety of scores is preferable to better identify those
tourist features of the city that can be improved.
V. C
This paper proposed a sentiment score of travel
destinations based on the content analysis of opinions shared
through the web. The case study about the city of Barcelona
shows that users are mainly focused on hotels and attractions,
and their opinions tend to avoid the use of very positive or
negative terms. According to the sensitivity analysis
implemented, it is better to overweigh very positive or
negative terms to achieve a higher variety of scores.
This work was supported by the Consejería de Economía,
Innovación, Ciencia y Empleo (Ministry of Economy,
Innovation, Science and Employment)viathe Research Project
P12-SEJ-328, by the Fundación Hergar (Hengar Foundation)
via the Research Project AIFH2014/005, and by the Programa
Estatal de Investigación, Desarrollo e Innovación Orientada a
los Retos de la Sociedad (State Programme of Research,
Development and Innovation Orientated to the Challenges of
Society) via the Research Project ECO2013-43856-R.
[1] K. Chon, “The role of destination image in tourism: A review and
discussion”, Tourist Review, Vol. 13, pp. 2-9, 1990.
[2] N. Tapachai and R. Waryszak, , “An examination of the role of
beneficial image in tourist destination selection”, Journal of Travel
Research, Vol. 39, Iss. 1, pp. 37-44, 2000.
[3] J. E. Bigné, M.I. Sánchez and J. Sánchez, “Tourism Image, Evaluation
Variables and After Purchase Behaviour: Inter-Relationship”, Tourism
Management, Vol. 22, pp. 607-616, 2001.
[4] N. Morgan, A. Pritchard, and R. Piggot, “New Zealand, 100 per cent
pure. The creation of a powerful niche destination brand”, Journal of
Brand Management, Vol. 9, Iss. 4-5, pp. 335-354, 2002.
[5] D. Stokes and W. Lomax, “Taking control of word-of-mouth marketing:
the case of an entrepreneurial hotelier”, Journal of Small Business and
Enterprise Develoment, Vol. 9, Iss. 4, pp. 349-57, 2002.
[6] J.S. Backman and S. Veldkamp, “Examination of the relationship
between service quality and user loyalty”, Journal of Park and
Recreation Administration, Vol. 13, Iss. 2, pp. 29-41, 1995.
[7] J.J. Cronin, M.K. Brady, and G.T.M Hult, “Assessing the effects of
quality, value and customer satisfaction on consumer behavioral
intnetions in service environments”, Journal of Retailing, Vol. 76, Iss. 2,
pp. 193-218, 2000.
[8] C. Ching-Fu, C. Fushian, “Experience quality, perceived value,
satisfaction and behavioral intentions for heritage tourists”, Vol. 31, Iss.
1, pp.29-35, 2010.
[9] R. Kashyap and D.C. Bojanic, “A structural analysis of value, quality,
and price perceptions of business and leisure travelers”, Journal of
Travel Research, Vol. 39, Iss. 1, pp. 45-53, 2000.
[10] J.F. Petrick, “Development of a multi-dimensional scale for measuring
the perceived value of a service”, Journal of Leisure Research, Vol. 34,
Iss. 2, pp.119-34, 2002.
[11] W. Kasper and M. Vela, “Sentiment Analysis for Hotel Reviews”,
Proccedings of the Computational Linguistics-Applications Conference,
pp 45-52, 2011.
[12] J. Hagel and G.A. Amstrong, Net Gain: Expanding Markets through
Virtual Communities. Boston, MA : Harvard Business School Press,
[13] Y. Bart, V. Shankar, F. Sultan and G. Urban, “Are the drivers and role of
online trust the same for all web sites and customers? A large-scale
exploratory empirical study”, Journal of Marketing, Vol. 69, Iss. 4,
pp.133-152, 2005.
[14] R.A. Shang, Y-C. Chen and H.J. Liao, “The value of participation in
virtual consumr communities on brand loyalty”, Internet Research, Vol.
16, Iss. 4, pp.398-418, 2006.
[15] C.M. Riddings, D. Gefen, and B. Arinze, "Some antecedents and effects
of trust in virtual communities", Journal of Strategic Information
Systems, Vol. 11, Iss. 3-4, pp. 271-295, 2002.
[16] L.V. Casaló, C. Flavián and M. Guinalíu, “ Fundaments of trust
management in the development of virtual communities”, Management
Research News, Vol. 3, Iss. 5, pp. 324-338, 2008.
[17] V.A. Zeithml, and M.J. Bitner, Service Marketing. International Edition.
Singapore: Mc-Graw Hill, 1996.
[18] S. Baloglu and M. Mangaloglu, “Tourism destination images of Turkey,
Egypt, Greece, and Italy as perceived by US-based Tour Operators and
Travel Agents”, Tourism Management, Vol. 22, pp. 1-9, 1999.
[19] A. Beerli and J.D. Martín, “Tourists’ Characteristics and the Perceived
Image of Tourist Destinations: A Quantitative Analysis-A Case of Study
of Lanzarote, Spain”, Tourism Management, Vol. 25, pp. 623-636, 2004.
[20] H. Kim and S.L. Richardson, “Motion Picture Impacts on Destination
Images”, Annals of Tourism Research, Vol.30, Iss. 1, pp.36-39, 2003.
[21] L. Andreu, J.E. Bigné and C. Cooper, “Projected and Perceived Image of
Spain as a Tourist Destination for British Travellers”, Journal of Travel
& Tourism Marketing, Vol. 9, Iss. 4, pp.47-67, 2000.
[22] N. Tapachai and R. Waryszak, “An Examination of the Role of
Beneficial Image in Tourist Destination Selection”, Journal of Travel
Research, Vol. 39, Iss. 1, pp.37-44, 2000.
[23] S. Baloglu and D. Bringer, “Affective Images of Tourism Destinations”,
Journal of Travel Research, Vol. 35, Iss. 4, pp. 11-15, 1997.
[24] J.E. Bigné and M.I. Sánchez, “Evaluación de la Imagen de Destinos
Turísticos: Una aplicación Metodológica a la Comunidad Valenciana”,
Revista Europea de Dirección y Economía de la Empresa, Vol. 10, Iss.
3, pp. 189-200, 2001.
[25] Y. Yoon and S. Kim, “An Assesment and Construct Validity of
Destination Image: A Use of Second-Order Factor Analysis”, Working
Paper, Virginia, USA, 2000.
[26] S. Pike and C. Ryan, “Destination Positioning Analysis Through a
Comparison of Cognitive, Affective, and Conative Perceptions”, Journal
of Travel Research, Vol. 42, Iss. 4, pp.333-342, 2004.
[27] D.J. Walmsley, and J.M. Jenkins, "Tourism Cognitive Mapping of
Unfamiliar Environments". Annals of Tourism Research, Vol. 10, Iss. 2,
pp. 268-286, 1992.
[28] S. Baloglu and M. Mangaloglu, "Tourism Destination Images of Turkey,
Egypt, Greece, and Italy as Perceived by US-Based Tour Operators and
Travel Agents", Tourism Management, Vol. 22, pp. 1-9, 2001.
[29] W.C. Gartner, “Image Formation Process”, Journal of Travel & Tourism
Marketing, Vol.2, Iss. 2-3, pp.191-215, 1989.
[30] J.T. Coshall,”Measurement of Tourists’ Images: The Repertory Grid
Approach”, Journal of Travel Research, Vol. 39, Iss. 1, pp. 85-89, 2000.
[31] H. Leemans, The multiform book: Using information in purchasing
hedonic products. Delft: Eburon, 1994
[32] R. Govers and F.M. Go, “Deconstructing Destination Image in the
Information Age”, Information Technology & Tourism, Vol. 6, pp. 13-
29, 2003.
[33] C.M. Echtner and J.R.B. Ritchie, “The Measuremen t of Destination
Image: An empirical Asseesment”, Journal of Travel Research, Vol. 31,
Iss. 4, pp. 3-13, 1993.
[34] J. Hanlan and S. Kelly, “Image formation, information sources and an
iconic Australian tourist destination”, Journal of Vacation Marketing,
Vol. 11, no. 2, 163-177, 2004.
[35] S. Aciar, “Mining context information from consumers’ reviews”, I
Proceedings of the Context-Aware Recommender Systems (CARS)
Workshop, 2009.
[36] C. Haruechaiyasak, A. Kongthon, P. Palingoon and C. Sangkeetrakarn,
“Constructing THai opinion mining resource: A case Study on hotel
reviews”, in Proceedings of the Eighth Workshop on Asian Language
Resources, 2010.
[37] V. Blazevic, W. Hammedi, I. Garnefeld, R. T. Rust, T. Keiningham, T.
W. Andreassen, N. Donthu and W. Carl, "Beyond traditional word-of-
mouth: An expanded model of customer-driven influence", Journal of
Service Management, Vol. 24 Iss. 3, pp. 294-313, 2013.
[38] F. J. Arenas-Márquez, M.R. Martínez-Torres and S.L. Toral, "Electronic
word-of-mouth communities from the perspective of social network
analysis", Technology Analysis & Strategic Management, doi
10.1080/09537325.2014.923565, 2014.
[39] R. A. King, P. Racherla and V. D. Bush, "What We Know and Don't
Know About Online Word-of-Mouth: A Review and Synthesis of the
Literature", Journal of Interactive Marketing, doi
10.1016/j.intmar.2014.02.001, 2014.
[40] A. Ortigosa, J. M. Martín and R. M. Carro, "Sentiment analysis in
Facebook and its application to e-learning", Computers in Human
Behavior, Vol. 31, pp. 527-541, 2014.
[41] F. A. Nielsen , "A new ANEW: Eva luation of word list for sentiment
analysis in microblogs", In Proc. of the ESW2011 Workshop on 'Making
sense of Microposts: Big Things come in small packages, no. 78 in
CEUR Workshop Proceedings, Heraklion, pp. 93-98, 2011.
... Recommendation based on emotion is one of the latest and the most popular research directions in the field of personalized recommendation (Orellana-Rodriguez et al. 2015;Musto et al. 2017;Fong et al. 2012;Gonzalez-Rodriguez et al. 2014;Vagliano et al. 2017) and Karypis (2001). Identifying and analyzing user behavior is its core strategy. ...
... Fong et al. (2012) proposed a semantic mining method to identify access patterns from usage logs by incorporating user emotions and behaviors based on selfreporting and behavioral tracking in a recommendation system. Gonzalez-Rodriguez et al. (2014) analyzed the content of user evaluations of tourist attractions based on emotion analysis techniques. In (Gonzalez-Rodriguez et al. 2014), the sentiment score is obtained to make recommendations by combining sensitivity analysis, tourist features and shared opinions. ...
... Gonzalez-Rodriguez et al. (2014) analyzed the content of user evaluations of tourist attractions based on emotion analysis techniques. In (Gonzalez-Rodriguez et al. 2014), the sentiment score is obtained to make recommendations by combining sensitivity analysis, tourist features and shared opinions. Vagliano et al. (2017) proposed a recommendation method based on the semantic annotation of entities that are noted in user reviews, and this approach treats the entities as candidate recommendations. ...
Full-text available
With the development of personalized recommendations, information overload has been alleviated. However, the sparsity of the user-item rating matrix and the weak transitivity of trust still affect the recommendation accuracy in complex social network environments. Additionally, collaborative filtering based on users is vulnerable to shilling attacks due to neighbor preference recommendation. With the objective of overcoming these problems, a collaborative filtering recommendation method based on trust and emotion is proposed in this paper. First, we employ a method based on explicit and implicit satisfaction to alleviate the sparsity problems. Second, we establish trust relationships among users using objective and subjective trust. Objective trust is determined by similarity of opinion, including rating similarity and preference similarity. Subjective trust is determined by familiarity among users based on six degrees of separation. Third, based on the trust relationship, a set of trusted neighbors is obtained for a target user. Next, to further exclude malicious users or attackers from the neighbors, the set is screened according to emotional consistency among users, which is mined from implicit user behavior information. Finally, based on the ratings of items by the screened trusted neighbors and the trust relationships among the target user and these neighbors, we can obtain a list of recommendations for the target user. The experimental results show that the proposed method can improve the recommendation accuracy in the case of data sparsity, effectively resist shilling attacks, and achieve higher recommendation accuracy for cold start users compared to other methods.
... Dictionary-based and corpus-based methods mainly achieve the building of sentiment lexicon, where corpus-based methods are further divided into statistical-based and semantics- based methods according to the specific techniques [9] . Gonzalez- Rodriguez et al. [10] analyzed sentiment orientations of travel- related information on social media via AFINN-111 [11] . The final sentiment score of a review was calculated based on the relative amounts of four types of sentiment words. ...
Online travel has developed dramatically during the past three years in China. This results in a large amount of unstructured data like tourism reviews from which it is hard to extract useful knowledge. In this paper, a DWWP system consisting of domain-specific new words detection (DW) and word propagation (WP) is presented. DW deals with the negligence of user-invented new words and converted sentiment words by means of AMI (Assembled Mutual Information). Inspired by social networks, the new method WP incorporates manually calibrated sentiment scores, semantic and statistical similarity information, which improves the quality of sentiment lexicon in comparison with existing data-driven methods. Experimental results show that DWWP improves seventeen percentage points compared with graph propagation and four percentage points compared with label propagation in terms of accuracy on Dataset I and Dataset II, respectively.
... As an active research field in Natural Language Processing (NLP), effective feature extraction is the basic requirement of text sentiment analysis. Dictionary-based and corpus-based methods mainly utilize the establishment of sentiment lexicon [5][6] to extract useful features for sentiment analysis of travel-relevant information on social media. And the key section of machine learning methods, such as Random Forests [7], SVM [8] and neural networks [9][10], is feature extraction and feature selection. ...
Full-text available
Sentiment analysis in tourism domain has drawn much attention in past few years, which calls for more precise sentiment word embedding method. The article proposes a kernel optimization function for sentiment word embedding. And the method aims at integrating the semantic information, statistics information and sentiment information and maintains the similarity between sentiment words in terms of sentiment orientation. The experiment result shows that the optimal sentiment vectors successfully extract the features in terms of sentiment information and the difference between concretization and abstraction of a sentiment words.
... Although some papers have analysed sentiment in a travel related context, they have focused on opinions about tourist destinations rather than sentiment about journeys (e.g., Gonzalez-Rodriguez, Martinez-Torres, & Toral, 2014). ...
Computer systems need to be able to react to stress in order to perform optimally on some tasks. This article describes TensiStrength, a system to detect the strength of stress and relaxation expressed in social media text messages. TensiStrength uses a lexical approach and a set of rules to detect direct and indirect expressions of stress or relaxation, particularly in the context of transportation. It is slightly more effective than a comparable sentiment analysis program, although their similar performances occur despite differences on almost half of the tweets gathered. The effectiveness of TensiStrength depends on the nature of the tweets classified, with tweets that are rich in stress-related terms being particularly problematic. Although generic machine learning methods can give better performance than TensiStrength overall, they exploit topic-related terms in a way that may be undesirable in practical applications and that may not work as well in more focused contexts. In conclusion, TensiStrength and generic machine learning approaches work well enough to be practical choices for intelligent applications that need to take advantage of stress information, and the decision about which to use depends on the nature of the texts analysed and the purpose of the task.
... The process of sentiment analysis involves text analytics, linguistics and accepted language processing to determine and dig subjective information from source materials. Sentiment analysis finds applications in various domains such as marketing, business and commerce (Jansen et al., 2009), healthcare (Chew and Eysenbach, 2010;Salathe and Khandelwal, 2011;Greaves et al., 2013), tourism and travel (Gonzalez-Rodriguez et al., 2014), and disaster management (Verma et al., 2011;Gao et al., 2011;Mandel et al., 2012). ...
Conference Paper
This short paper describes a sentiment analysis system for micro-post data that includes analysis of tweets from Twitter and Short Messaging Service (SMS) text messages. We discuss our system that makes use of Word Sense Disambiguation techniques in sentiment analysis at the message level, where the entire tweet or SMS text was analysed to determine its dominant sentiment. Previous work done in the area of Word Sense Disambiguation does not throw light on its influence on the analysis of social-media text and micropost data, which is what our work aims to achieve. Our experiments show that the use of Word Sense Disambiguation alone has resulted in an improved sentiment analysis system that outperforms systems built without incorporating Word Sense Disambiguation.
Purpose – The importance of virtual communities is growing day by day, since consumers are increasingly using online tools to share ideas and contact fellow consumers. For this reason, the purpose of this paper, which is based on relational capital theory, is to analyze what factors determine the consumer commitment to a virtual community. Design/methodology/approach – A positive effect of trust in a virtual community on commitment to the community is suggested. In addition, some antecedents of trust placed in a virtual community are proposed. After the validations of measurement scales the hypotheses are contrasted through structural modelling. Findings – The data show that trust placed in a virtual community has a positive and significant effect on consumers' commitment to that virtual community. Additionally, a general disposition to trust, a greater familiarity with the community and a stronger norm of reciprocity in communication in the community may increase the level of trust placed in a virtual community. Research limitations/implications – Data were collected thanks to a web survey using Spanish-speaking members of free software virtual communities. Thus, it would be interesting to carry out a new validation of the model using other types of virtual communities and a wider sample of consumers, particularly in terms of different nationalities. Practical implications – This study has shown how managers may foster trust and commitment toward a virtual community in order to ensure the community's success and survival in the long term; that is, the ongoing participation in the community. Originality/value – Most of the works that are focused on virtual communities have been conducted at the conceptual level. Thus, with the aim of progressing this topic, the study analyzes empirically the precursors of trust and commitment to a virtual community.
Full-text available
User reviews and comments on hotels on the web are an important information source in travel planning. Therefore, knowing about these comments is important for quality control to the hotel management, too. We present a system that collects such comments from the web and creates classified and struc-tured overviews of such comments and facilitates access to that information.
Conference Paper
Full-text available
User reviews and comments on hotels on the web are an important information source in travel planning. Therefore, knowing about these comments is important for quality control to the hotel management, too. We present a system that collects such comments from the web and creates classified and struc-tured overviews of such comments and facilitates access to that information.
Full-text available
This paper is focused on the identification of influencers that can have an important impact over the decision-making of other users. For this purpose, a popular electronic word-of-mouth community like has been modelled as a social network. Using social network analysis techniques, the existence of influencers is justified by the power law distribution of user participation, and then they are identified using their topological features within the social network. The obtained results reveal that influencers are not determined by the number of performed reviews, but by the variety or scope of their performed reviews and their central position in the consumer network. The main contribution of this research is the identification of influencers based on the participation features of community users. As a difference to other studies, results are not based on surveys or opinions, but on the trace users leave when they post opinions, comments or scores.
Understanding how tourism destination image (DI) is formed may assist destination promoters in developing appropriate images for their destination services and products. This study theoretically develops and empirically tests a measurement model of tourism destination image. Through a second order factor analysis in LISREL, the results confirm that destination image can be operationalized as a second order factor model that includes an affective image (AI) and a cognitive image (CI). These constructs are governed by a second order construct of tourism destination image. Interestingly, the results of this study suggest that the affective construct has more impact on building destination image than does the cognitive construct. Finally, theoretical implications and practical suggestions are discussed.
Electronic word-of mouth (eWOM) has attracted considerable interest from researchers in the past decade. Although the extant research has helped us to develop a good understanding of a number of the issues pertaining to eWOM, several research and managerial questions remain. Furthermore, no attempt has been made to consolidate and synthesize this stream of research. With consumers' increasing reliance on online retailing and information seeking, as well as the continued growth of social media, the importance of eWOM cannot be overstated. Based on a systematic review of 190 studies, we conduct a multi-dimensional analysis of eWOM communication. We present the key issues in current and emerging literature and propose important questions for future research.
This paper presents a new method for sentiment analysis in Facebook that, starting from messages written by users, supports: (i) to extract information about the users' sentiment polarity (positive, neutral or negative), as transmitted in the messages they write; and (ii) to model the users' usual sentiment polarity and to detect significant emotional changes. We have implemented this method in SentBuk, a Facebook application also presented in this paper. SentBuk retrieves messages written by users in Facebook and classifies them according to their polarity, showing the results to the users through an interactive interface. It also supports emotional change detection, friend's emotion finding, user classification according to their messages, and statistics, among others. The classification method implemented in SentBuk follows a hybrid approach: it combines lexical-based and machine-learning techniques. The results obtained through this approach show that it is feasible to perform sentiment analysis in Facebook with high accuracy (83.27%). In the context of e-learning, it is very useful to have information about the users' sentiments available. On one hand, this information can be used by adaptive e-learning systems to support personalized learning, by considering the user's emotional state when recommending him/her the most suitable activities to be tackled at each time. On the other hand, the students' sentiments towards a course can serve as feedback for teachers, especially in the case of online learning, where face-to-face contact is less frequent. The usefulness of this work in the context of e-learning, both for teachers and for adaptive systems, is described too.
The destination image and positioning studies in tourism have been limited to those dealing with the image's perceptual or cognitive component. This study examined the applicability of Russel and his colleagues' proposed affective space structure to large-scale environments (i. e., tourism destination countries) as well as its potential as a positioning structure to study affective images of tourism destinations. The multidimensional scaling analysis of 11 Mediterranean countries along with proposed affective space structure indicated that Russel and his colleagues' proposed affective space can also be applied to places that are not perceived directly. It also showed potential for studying the affective image positioning of tourism destinations. The article concludes with some theoretical and practical implications and future research areas regarding tourism destination images.