The Tourism Destination Image (TDI) has been usually studied by conducting tests with individuals visiting the destination for which the TDI is being constructed. This paper explores the structure of the cognitive, affective and conative components of tourism destination image using social data. Then, it reports the progress in the implementation of a pilot system for the study of TDI using data extracted from social media which is being built using the SLOD-BI semantic infrastructure for a use case involving Vilafamés, a rural tourist destination in Eastern Spain.
Although there are a variety of different interpretations for
the concept of Tourism Destination Image (TDI), it is
commonly recognized that the TDI is “the sum of beliefs, ideas,
and impressions that a person has of a destination” [
In the past, the TDI was studied as formed only by
cognitive components, but nowadays it is agreed that the TDI is a
multidimensional overall impression formed by distinctly
different but interrelated components, namely cognitive,
affective, and conative [
In the literature, the TDI is explored by having visitors fill
in questionnaires, usually when they are leaving the tourist
destination [
which it is possible to get tourists opinions. b. Determine when the opinion of a user is re
lated to an attribute under study. c. Define a measurement scale for each attrib
ute.
4. Gather relevant social data for the case of study.
5. Develop a tourism sentiment analysis process for the
concepts related to the affective component [
The next sections present the first steps of some of the points in this plan. Next section presents the infrastructure used to get and gather the social data and outlines how the sentiment analysis of the opinions is going to be done. Section 3 presents the social networks that are going to be used, and the most used attributes to measure each component, and how some of them can be gathered from the social networks. Finally, we conclude with an outline of future work. 2
We will adopt the SLOD-BI infrastructure [
SLOD-BI incorporated facilities for the large-scale processing of opinion data extracted from social media. It provides a data model based on BI patterns, which abstract the data required for analysis of social media into a set of generic facts and dimensions. The key patterns for our use case are Post facts, which contain information about a textual posting in an online source that may contain opinion information. Social facts group contextual information about the post facts that will help assess its impact, such as the number of items the opinion has been shared, or the number followers of the author of the post. Finally, opinion facts contain the results of an automated sentiment analysis computation, including the object and facets that are being opined about, and the polarity of the opinion (positive, negative or neutral). Figure 1 shows the relationship between these social facts, and how the SLOD-BI model also allows the integration between social facts and corporate data, which includes internal company data and external resources such as relevant economic indicators.
Populating a data model such as this requires strong data
integration capabilities. SLOD-BI adopts the LOD (Linked
Open Data) paradigm [
Figure 3 shows the functional architecture of SLOD-BI, illustrating how the source data is progressively processed, with the final goal of building decision support systems based on analytical tools, predictive models and exploratory interfaces. 3
As a first step, we contacted interviews with both tourismrelated public and private stakeholders (tourism officers, owners of tourism-oriented companies, and local government authorities related with tourism management.). This allowed us to gather necessary data to get information about market positioning, and future expectations.
We did a preliminary study to select the relevant sources of information. We included: 1 2
General social networks on which there is a significant presence of local tourism-related organizations. These are Facebook and Twitter.
Tourism-specific social networks, extracted from the
social media directory at [
The results of this analysis are presented in Table 1. The selected social data sources were Facebook, Twitter, Minube, TripAdvisor and Yelp.
After selecting the data sources, we performed an exploratory analysis of them. First, we studied the hashtags related with the small village (Figure 4), and then a more exhaustive analysis of the type of information found in each source has been conducted. For instance, Figure 5 shows an example of the information a tourist can found when asking about accommodation facilities in Vilafamés.
From a literature review, and given the case of study under consideration, we have chosen the items shown in Table 2 for studying the cognitive, affective and conative dimensions of Vilafamés’s TDI.
The three dimensions under study are related, and each one of them contribute to the formation of a global image that is considered to be greater than the sum of its parts, and this is used by the consumer to simplify the task of decision-making.
On the other hand, as stated in [
As first steps, for studying the dimensions of the village we have collected opinions of users as the ones shown in Figure 5 and we have carry on a sentiment analysis of them. The result of this task determines: the language of the opinion, the polarity of the general opinion, and the facets detected. We are yet working on the detection of the facets or features about what the opinion is talking about. The facets study is still under development in order to determine which will be the final facets considered when opining about a TDI and how determine the polarity (positive, neutral or negative) of all of them. The facets have to be defined taking into account the items of the dimensions. For instance, below we present the result of this sentiment analysis task done for some sentences extracted from TripAdvisor about Vilafamés. For all of them the system has determined that they are written in Spanish, in the sentences the facets are highlighted using italics. For the two first sentences the analysis returns a positive polarity and for the last one a neutral one.
“Escapada de relax. Trato muy agradable tanto de la propietaria como del servicio, una señora brasileña muy simpatica. La habitacion muy amplia y limpia. Lo mejor el desayuno "perfecto". El pueblo precioso e idilico! Volveremos....”.
“Acabamos de estar en el hotel la habitacion amplia y todo muy limpio un poco cara precio / calidad anoche cenamos en el restaurante y lo mismo caro precio / calidad , pero el desayuno ha sido una pasada fantastico de todo dulces y salado en cantidad y calidad y un 10 a los bizcochos caseros yo volveria a ir solo por volver a desayunar”.
“Pasamos una noche en el Hotel el Rullo y quedamos muy satisfechos. Las habitaciones son cómodas y estaban limpias. No podíamos pedir más. El único inconveniente que tuvimos fue el desayuno. Es un desayuno fijo, muy bueno todo, pero tardaron 30 minutos de reloj en servirnos el café, el zumo y las tostas en la mesa. Sólo había una chica sirviendo los desayunos más quien hubiese en la cocina, pero claro estaba que no daban a basto. Deberían mejorarlo, sobretodo en épocas con más afluencia de huéspedes. Por lo demás, nos fuimos contentos.” We have started the research methodology to determine the Tourist Destination Image (TDI) of a small village, Vilafamés. First we have determined the three dimensions to study, the cognitive, affective and conative dimensions, their relations and the items to be studied for each one of them. The social networks to be used for this analysis have been also determined and the SLOD-BI architecture has been selected for developing the whole system. The initial sentiment analysis of the opinion has been presented too.
Now, there is still a lot of work to be done in order to get the final TDI of Vilafamés. We have to finish yet the facets definition from the opinions, and it is necessary to associate also a sentiment for each facet. Then it is necessary to combine all the information gathered from the different social networks for each facet and dimension, to allow analysis on the global data. Also it is interesting to detect communities in the network connection graph and to determine which it is the most suitable way to visualize all the data in order to be useful for tourism decision-making.
This work has been partially supported by the Andalusian Regional Ministry of Economy (project SIMON TIc-8052), the Spanish Ministry of Economy and Competitiveness (project TIN2014-55335R), Generalitat Valenciana (project GVA/2015/102) and Universitat Jaume I (project P11B201329).