This paper describes main modern tendencies for the design and development of e-tourism recommender systems with sentiment analysis of user generated content in social media. Main goal is to systematize and summarize knowledge about the possibilities of using tourist's user reviews in social media as a type of e-tourism big data for mobile e-tourism recommender systems. In particular, to analyze the sources and types of tourist feedback data, messages and comments generated by the tourist with his gadget that can be related to etourism big data. Developing efficient tools for e-tourism user comments and feedback in social media, combining big data technologies, NLP and smartphone services advantages, can provide e-tourism recommender systems with new better ways of creating more personalized recommendations.
There is an increasing interest for the use of content created by consumers of hospitality and tourism services, in particular on social networks and video hosting. Thus, the structure and dynamics of tourists' preferences can be tracked and analyzed, information about the image and reputation of the tourist product can be received, as well as about the behavior of tourists themselves when traveling. The feedback received from the tourist is not only useful for business, but also can be used by recommender applications as one of the sources for estimation of the alternative item. Two popular eadvice websites Booking and TripAdvisor host users’ opinions since decades. But they are very much moderated. Also not every user leaves feedback on tourism-related review platforms. But every user has a profile in one or more social networks. And there he publishes different aspects of his life, tourist experience included.
Copyright © 2020 for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
E-Tourism Recommender Systems
Recommender systems are a class of intelligent information retrieval systems
designed to filter out, in a abundance of information resources, exactly the instances of
data that best meet the interests of a particular user [
E-tourism recommender systems can be classified according to different
characteristics, such as: architecture, information technology platform, target audience,
methods used for generating recommendations, main tasks to be solved, etc [
Knowledge base
Methodology d e t n e i r o b e W g n i n n a l P ) C P ( p o t k s e D e d i u G
e l i b o M n o i t a g i v a N c i t a t S n o i t a z i m i t p O l o r t n o c t s o C c i m a n y
D l e v a r t f ts o l s su isy re l a n A t s i r u o T s t s ir u o tf o p u o r G
t n e g A t- d ex te t en on ir C o t- d en te t en on ir C o d i r b y H The basis for successful operation for mobile e-tourism recommender systems is a complete, well-timed and correct data processing. There are certain peculiarities in dealing with input and output data and they should be taken into account when designing tourist recommender systems, in particular mobile ones. There are three main sources of input data for a mobile e-tourism recommender system:
1) The user as an informational source – generates queries, leaves feedback, disseminates messages about himself on social networks. All smart tourism technologies nowadays act in the paradigms "tourist as a sensor" and "every tourist is an expert".
User info profile (including feedback, reviews and comments)
Bluetooth and other sensors data 2) The gadget of the user itself - information about the external background of the tourist, contextual data, appearance or disappearance from the operational space of various obstacles, etc.
3) Internet content and internet of things - this is data from referential resources, both tourist and external, work schedules, lists of tourist places and establishments, public transport timetables, etc., including web search data, user net surfing history and online booking data, and more.
user Internet of things
Tourist services
The gadget user is a powerful source of information for e-tourism software products.
This data belongs to the e-tourism big data category [
In particular, the text message that a gatget user leaves as a review of a tourism
product consumed has changed [
1. The user response has become shorter. First, there are limits to the number of
characters for a single message in different systems, such as Twitter; screen size of
the smartphone – there is an unspoken rule "what wasn’t fit on one screen will not be
read" [
2. The space for posting reviews has also changed. Traditionally, users have left
posts on specialized sites, travel forums, travel agency blogs, and more [
3. The structure of the comment has also changed: text is now being supplemented
or even partially replaced by graphic, audio and video content. Emoji, stickers,
animated elements convey the emotional tinge of user feedback. Video stories and live
streams may contain text captions and subtitles to increase the content of the response
[
4. The user may not plan and prepare the response text in advance, his story may be devoted to a completely different topic, and his own impression of the consumed tourist product will "slip through" among other things. Such reviews are the hardest to follow, but they also create a reputation for the tourist product [15].
5. Using a hashtag for text and geo-positioning for images and videos allows you to uniquely identify the tourism product [16]. Making it easier to find and analyze data.
6. Option of personalized feedback from the author of the review. From the official owner’s profile of the tourist product can be added in response to the tourist post a gratitude for the positive feedback or an apology in case of complaints [17-18]. In this way, thanks to the social media space, the product seller is able to reach the customer's information territory and attract his (and his social environment) attention. It is also possible to supplement user profile of the recommendation application with new review facts.
7. The language used by the tourist: in the forums and official pages of the tourist objective (classic space for creating reviews), as a rule, one language is used, or in the case of regional information resources, two: English and the language of the region [19]. The social media space enables the user to express his or her thoughts in the language best suited to them [20]. That is, it is likely that the tourism product provided by one country will have reviews in five, ten or more different languages. Which complicates the analysis of the text.
Therefore, travel product reviews need to be collected not only on specialized resources, but also increasingly in the social media space. Analyzing the sentimental content of such reviews is complicated, first, by multilingualism and, second, by the presence of such graphic elements as emoji and stickers. Consumer feedback now needs to be maintained on users territory – on social media.
Accordingly, the analysis of the sentimental filling of tourist feedback on tourist products is not only a source of data for mobile e-tourism recommender systems, but also transforms from the classic text-mining task to the task of analyzing big data, not only text [21].
Finding and retrieving useful information from user reviews of a tourism product in the social media space poses a number of challenges to developers of recommender applications. In particular:
1. How to properly treat sentimental tint of an emoji in reviews? Is negative content related to the mood of the user, the weather, the day in general or the quality of the tourist product consumed this day? Is it possible to use for comparison as a description of previous bad experiences with another product? Should emoji be considered equivalent to keywords in reviews?
2. How situational and implicit (no hashtag and location) reviews can be well tracked and consolidated? 3. How to effectively extract text content from photos, videos and audio messages? 4. Should the publication a tourist photo or video related to the tourist product but without supplementing the text message be considered as a feedback and how to classify it: as positive, negative or neutral?
These and other problems need to be solved to create efficient sentiment analysis technologies for mobile e-tourism recommender systems. 4
Using Sentiment Text Analysis for E-Tourism Technologies Natural Language Processing (NLP) is a field of Computer Science that studies the use of automatic ways to process natural language. Sentiment text analysis is a fast growing element of NLP [22]. Automatic processing of e-tourism text data due to the large amount of content generated by users every minute is becoming more complicated. The keys to solve it lie in combining big data technologies, NLP and smartphone services advantages [23].
Different domains and types of texts have different information extraction requirements and thus require different NLP tasks and tools [24]. Developing efficient tools for sentiment analysis of specific type of text – e-tourism user comments and feedback in social media can provide e-tourism recommender systems with new better ways of creating more personalized recommendations.
There are researches and discussions on the mechanisms behind reviewing tourists behavior and it’s connection with the data of the reputation sites, hotels, attractions and destinations have online, and how this affects tourist behavior and purchasing decisions. Social media feedback data bring new context and new challenges to this topic. But also they bring new perspectives and resources. 5
Sentiment Text Analysis of E-Tourism User Reviews In the first stage, a list of key components of the response was compiled: keywords, hashtags, emoji, and the order of punctuation was drawn. The keywords were divided into classes, as well as Ukrainian, English and Romanian, since these three languages are used by tourists to provide feedback on Bukovina tourist services, as shown in Tables 1, 2 and 3. Since Protégé cannot write hashtags via "#", we wrote them using the letter "h". The hashtags were divided into "Very Positive", "Positive" and "Neutral" as well as being Ukrainian, English and Romanian as shown in the tables. 4, 5 and. 6. Дуже позитивні h_дуже h_дужесмачно h_дужевесело a
Foarte pozitive h_foartegustos h_foartevesel h_foarteieftin h_foarte_frumos h_foarte_tare h_foarte_amuzant h_foartefoarte h_vesel h_divin h_perfect h_perfectazi h_perfectadimineata h_perfecta_zi h_perfectaseara h_perfecta_dimineat h_pefecta_seara h_мандрівки h_мандрівник h_мандрівниця h_мандрівники h_мандруйдешевше h_мандруй_сміливо h_мандруємоукраїною h_мандрівникиукраїною h_мандруюукраїною h_мандриподорожі h_мандруватилегко h_мандруй_з_нами h_мандруй_активно h_подорож h_подорожі h_подорожіукраїною h_подорожуйукраїною h_подорожі_україною h_подорожуємо h_подорожуйзнами h_подорожувати h_подорожуєморазом h_подорожуй_україною h_подорожуючиукраїною Since we can't add emoji to Protégé, we wrote them through the letter "e". Emoji were divided into "Very Positive," "Positive," "Neutral," "Negative," and "Very Negative." Table 7.
Emoji Punctuation marks are used to denote such a dismemberment of a written language that cannot be transmitted either by morphological means or by the order of the words in the sentence.
An exclamation point (!) Is a punctuation mark that is placed at the end of a sentence to express outrage, a call for strong feelings, anxiety, and more. It can also be doubled, tripled or used many times to express greater expression and emotionality in grammatical abuse.
Question mark (?) Is a punctuation mark, usually placed at the end of a sentence to express a question or doubt.
In user reviews, punctuation marks such as question mark (?) And exclamation mark (!) Are very common, they can be for positive feedback as well as negative feedback, it all depends on the words found before punctuation marks, positive keywords or negative .
Users use punctuation to express or displease tourist services. If after a positive keyword there are three exclamation marks then the keyword refers to very positive feedback, but if after a positive keyword there are three question marks then the keyword refers to very negative feedback. A single exclamation mark after neutral keywords means that the keyword refers to positive responses, but if one question mark after a negative keyword means the keyword refers to neutral responses. For example, a user posted the following comment: Like it! this keyword is not a positive but a very positive one, because there are three exclamation points after it, or the user left a "Dear !!!" this keyword refers not to negative but very negative feedback, or the user left a response: "Why is it so expensive?", the keyword here is "expensive", since after the keyword one question mark, the response refers to neutral feedback.
According to the keyword tables, hashtags and emoji built a hierarchy of ontology classes and subclasses with Protege software. The classes in Protege are displayed as a class hierarchy (Class Hierarchy). Initially, they created base classes according to the hierarchy. Instances were created for each class as shown in Figure 4. The ontology properties were created, corresponding to the areas of definition and areas of value of the hierarchical ontology. Figure 5.
Conclusions
This study is an attempt to systematize and summarize knowledge about the possibilities of using tourist’s user reviews in social media as a type of e-tourism big data for mobile e-tourism recommender systems. In particular, to analyze the sources and types of tourist feedback data, messages and comments generated by the tourist with his gadget, that can be related to e-tourism big data. 14. Rossetti, M., Stella, F., & Zanker, M.: Analyzing user reviews in tourism with topic models. Information Technology & Tourism, 16(1), 5–21. (2015). doi:10.1007/s40558-0150035-y 15. Krawczyk, M., & Xiang, Z.: Perceptual mapping of hotel brands using online reviews: a text analytics approach. Information Technology & Tourism, 16(1), 23–43. (2015). doi:10.1007/s40558-015-0033-0 16. García-Pablos, A., Cuadros, M., & Linaza, M. T.: Automatic analysis of textual hotel reviews. Information Technology & Tourism, 16(1), 45–69. (2015). doi:10.1007/s40558015-0047-7 17. Francalanci, C., & Hussain, A.: Discovering social influencers with network visualization: evidence from the tourism domain. Information Technology & Tourism, 16(1), 103–125. (2015). doi:10.1007/s40558-015-0030-3 18. Tao, M., Nawaz, M. Z., Nawaz, S., Butt, A. H., & Ahmad, H.: Users’ acceptance of innovative mobile hotel booking trends: UK vs. PRC. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0123-x 19. Fazzolari, M., & Petrocchi, M.: A study on online travel reviews through intelligent data analysis. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0121-z 20. Zhang, W., & Fesenmaier, D. R.: Assessing emotions in online stories: comparing selfreport and text-based approaches. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0122-y 21. Ghahramani, L., Khalilzadeh, J., & KC, B.: Tour guides’ communication ecosystems: an inferential social network analysis approach. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0114-y 22. Van der Zee, E., & Bertocchi, D.: Finding patterns in urban tourist behaviour: a social network analysis approach based on TripAdvisor reviews. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0128-5 23. Al-Ghossein, M., Abdessalem, T., & Barré, A.: Open data in the hotel industry: leveraging forthcoming events for hotel recommendation. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0119-6 24. Qi, S., Wong, C. U. I., Chen, N., Rong, J., & Du, J.: Profiling Macau cultural tourists by using user-generated content from online social media. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0120-0 25. . Chan, I. C. C., & Law, R.: Tanja Schneider, Karin Eli, Catherine Dolan, and Stanley Ulijaszek (editors): Digital food activism. Information Technology & Tourism. (2018). doi:10.1007/s40558-018-0117-8