Recommending Crowdsourced Trips on wOndary Linus W. Dietz Achim Weimert Technical University of Munich wOndary LTD Garching, Germany London, United Kingdom linus.dietz@tum.de achim@wondary.com ABSTRACT and further discuss avenues to improve the current basic system in Making recommendations for tourist trips is a challenging task due Section 4. We conclude this paper in Section 5. to the intrinsic complexity of the domain. The characterization of itineraries is non-trivial, because there is a lack of open destina- 2 TRAVEL RECOMMENDATION FOR tion databases such as regions, islands, cities or attractions that INDEPENDENT TRAVELERS would help to understand the characteristics of destinations within wOndary is a platform that allows users to save, organize and share a trip. For this purpose, we present wOndary, which supports the details about their trips. The platform helps with the structuring of planning and sharing of worldwide trips based on crowdsourcing. personal itineraries, enables collaboration between group travel- We sidestep item discovery and routing challenges of the tourist ers, and encourages the publishing of personal itineraries so that trip design problem by performing content-based recommendation others can reuse and customize these crowdsourced trips for their by facilitating a novel structured representation of itineraries. We own purposes. wOndary currently focuses on young urbans (23–30 share our experiences in the establishment of the core model for year olds) that strive for unique experiences during independently our travel recommender system and discuss future developments. planned trips. The user journey on wOndary reflects the travel micro-moments KEYWORDS as defined by Google as “dreaming, planning, booking, and experi- recommender systems, user modeling, crowdsourcing, explaining encing” [11]. When users dream of going away, they browse crowd- recommendations, critiquing sourced itineraries on wOndary or read travel-related stories. On- line travel media, such as travel blogs can include wOndary’s widget 1 INTRODUCTION to refer users to unique itineraries that have been created by other Independent travel planning is very complex. Today’s connected travelers. In the planning phase the users save activities or copy world offers a myriad of choices of where to travel to, and there itineraries to quickly create their own, customized trip. The users is unlimited information based on which one can make a decision. can search for specific locations and activities on and off the plat- wOndary1 has developed a platform for independent travelers to form and collaborate with their co-travelers. By synchronizing the plan their trips. Initially, it started as a planning tool to create wOndary itinerary to the calendar app on their phone, the trip info personal itineraries that can be shared privately with friends and becomes available when a user is offline to experience the foreign co-travelers or that can be published as a public trip on the platform. culture, but can be adapted at any time if there is Internet connectiv- In this paper, we describe how we transition the wOndary platform ity. Once the users return from their trip, they can privately share to a personalized recommender system for crowdsourced trips and their itinerary with friends and colleagues or decide to publish it to describe the future potential of this work. all other users within the platform. Our proposed solution involves the following contributions. We wOndary features a web application that is currently available present a data model for structuring trips into blocks that are both in open beta on https://wondary.com. It is implemented as a single- useful for users and for segmenting trips. Furthermore, we present page-application that runs on the Google Cloud Platform, and there- an attractions categorization that enables content-based recommen- fore, works in web browsers on all types of devices. dations via implicitly elicited preference. Utilizing this novel data model for structured itineraries, we provide recommendations for 2.1 Finding Inspiration with the “Explore” complete trips, and for parts of trips, i.e., blocks. The approach was Page driven by the following research questions (RQs): The users need a structured way to access the growing number of RQ 1: What is a suitable recommendation model that masters the crowdsourced itineraries. To answer our first research question, complexities of travel and enables future innovation regard- wOndary provides a location-based “Explore” feature that allows ing the user experience within wOndary? querying a location, and filtering by geographical bounds and at- RQ 2 How can crowdsourced trips be structured and characterized tributes, such as trip duration. Filtering is possible by adjusting the to enable content-based recommendations? trip duration (72 hours, 1 week, or 2+weeks), the season, and the RQ 3: How can user preferences be elicited without requiring query area by adjusting the map excerpt. We chose this visual rep- much effort by the user? resentation for the recommendations because a complex domain, In the following section, we describe wOndary and the core of like global travel, requires an intuitive user interface instead of the novel content-based travel recommender. Then, in Section 3 a simple list. Therefore, matching items are displayed on a map we survey prior literature on personalized travel recommendation and as a list, where their ranking depends on the distance to the 1 https://wondary.com queried location constituting the baseline for future improvements. RecTour 2018, October 7th, 2018, Vancouver, Canada. 13 Copyright held by the author(s). Figure 1: Explore Page, https://wondary.com/explore The recommended items are both full itineraries as well as blocks, approach planning. Additionally, it allows the normalization of trips as defined below. As can be seen in Figure 1, the interface allows spanning longer periods of time (several weeks or months) into users to zoom into a geographical region. The users can also view portions that are transferable between trips of different travelers. high-level information about itineraries and focus on the ones they wOndary heavily relies on blocks, not only when recommending would like to see further details about, or they can copy them as a items but also when presenting structured information about trips basis for their own customized trip. to users. With an increasing number of trips being published on the plat- The next lower level of the data model is the day, consisting form, it has become difficult for users to identify itineraries that of three types of entries: transportation, lodging, and activities. fit their travel requirements. For example, itineraries are diverse Having a good overview of how to get from one place to another in terms of included activities, and a user who loves sightseeing and where to stay overnight is essential for planning travel, whereas, may not be interested in a trip that features primarily beaches or instead, travelers define their trips based on the attractions they a multi-day hike through the mountains. Additionally, the users visit during the day. Currently, the users can input the attractions expect websites to support them in finding relevant content. Last, using venues from Google Places to ensure that they actually exist; the number of trips for a popular region make it tedious for users to typos are corrected, and duplicates are eliminated. Furthermore, review all itineraries. Therefore, recommendations are playing an the Google Places service provides further information, such as an increasingly important role in wOndary’s Explore feature because image, opening hours, or ratings. showing relevant content to the user improves their engagement To perform content-based recommendations, it is necessary to and general satisfaction with the application. classify items and the users into some meaningful categories. There- fore, our answer to RQ2 is the aforementioned data model using 2.2 A Data Model for Structured Itineraries the five categories listed below, which are influenced by the target audience of the platform and the available attraction information. wOndary’s data model for trips is based on insights from the domain. We compiled them based on an analysis of the platform’s trips When travelers plan their trips, they often think of destinations, combined with our expert knowledge on individual travel. e.g., cities that they want to connectively visit. For example, a trip to Italy would start with several days in Rome, then, a day in Florence, • Food Mainly comprises restaurants and cafés, but also gro- visiting friends in Bologna over the weekend, and finish with three cery stores and food markets. more days in Venice. To capture this, wOndary structures trips into • Culture Describes activities and places with cultural or his- blocks. A block acts as a descriptor of a partial trip that has a dura- torical attributes. For example, museums, galleries, churches tion of one or more consecutive days and links to a location. Thus, and theaters fall under this category. trips are modeled as a sequence of one or more blocks. This structure • Nightlife Categorizes places that are commonly related to was designed using user feedback and matches the way travelers nightlife such as bars, night markets, and jazz clubs. RecTour 2018, October 7th, 2018, Vancouver, Canada. 14 Copyright held by the author(s). food culture nightlife outdoor travel 50 100 40 Frequency 30 Frequency 50 20 0 10 0 50 100 150 Trip 0 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 Category food culture nightlife outdoor travel Trip Figure 2: Frequencies of Categories per Trip Figure 3: Classification of Trips per Category • Outdoor Includes attributes associated with natural scenery or outdoor activities, such as parks, nature preserves, beaches, mountains and trails. For example, hospitals are not mapped to any of our categories, • Transport & Travel Consists of travel-related attractions, because we argue that they are not relevant for planning a trip. such as ferries, train stations and airports. This indicates that Figure 2 shows the distribution of categories of a representative a relevant portion of the day is spent on transportation and sample of 150 trips from wOndary based on the top trips according that the transfer itself is an attraction. to user interactions. A closer look into the distribution of the cate- gories in Figure 3 shows that most venues are categorized into the 2.3 Content-Based Travel Recommendations Outdoor category, whereas Nightlife is the least frequent. Having classified the items, it is also necessary to know the To categorize the attractions, we query the Google Places types2 and user’s preferences to do content-based recommendation. The de- directly map them into our five categories. However, the returned fault method would be to explicitly ask the user to indicate her place types are not primarily meant for travelers. For example, the preferences regarding the five travel categories, e.g., on a scale from type query for the Colosseum of Rome, Italy returns: 1 (not interesting) to 5 (highly interesting). However, this would "types": ["point_of_interest", "establishment"] require a manual interaction, which we can avoid by using synergy While these types are not totally off mark, the information is effects from the categorization of attractions. To answer RQ3, we insufficient to categorize this monument into one of our categories. aggregated all attractions from a user’s saved trips to create a user Therefore, we augment the types from Google with an additional preference profile. While this can be refined further with more de- lookup of the attraction via the Foursquare API to allow one at- tailed click stream data, it is a straightforward metric for classifying traction to be a member of several categories. Foursquare has a user travel preferences within wOndary. rich hierarchical region categorization3 with 923 categories that The actual ranking for the recommendations is performed by are organized in a tree to model specialized subcategories. To lo- calculating the cosine similarity using the five dimensional vector cate a Google Place on Foursquare, we performed a query by name of distinctive travel interests. Here we exploit the structure of our using the exact location. By doing a bulk comparison, we found data model to recommend complete trips and trip parts, i.e., blocks that most attractions also exist in Foursquare, except for political or specific days. Currently, the Explore page features trips and entities, such as city names. Conveniently, due to the bounded local blocks as recommendations. In the first step, the system filters out search, the first result for Foursquare was the correct result for the all trips that are not within the bounds of the map or do not match corresponding Google Place. Recalling our example, we found that the temporal filters (see Figure 1). When a trip is only partially in Colosseum was categorized as a “Historic Site”, which is within the the query region, the blocks within the area will be included. Then, “Arts & Entertainment” category of Foursquare. Using static mapping all past trips of the user are removed because we assume they are of all Google types and Foursquare categories, we can determine the not of interest for future travel plans. Trips and blocks as ranked wOndary categorization. The Colosseum would be categorized into by the cosine similarity with respect to the user profile and also Culture because the Google types (‘point_of_interest’ and ‘establish- listed left of the map. To keep clarity in the interface, only the top ment’) are not part of the mapping, whereas a ‘Historic Site’ maps 30 items are displayed. to the Culture category. One attraction can have several wOndary For new users that have not yet copied any trips, the content- categories; however, not all venue types are relevant for travelers. based recommender cannot compute a ranking for the trips. There- 2 https://developers.google.com/places/web-service/supported_types fore, the trips displayed on the Explore page will be ranked by the 3 https://developer.foursquare.com/docs/resources/categories geographic distance to the center of the map. RecTour 2018, October 7th, 2018, Vancouver, Canada. 15 Copyright held by the author(s). 3 STATE OF THE ART OF PERSONALIZED commercial data set of 30,000 tourist destinations classified along TRAVEL RECOMMENDATION 27 motivational and 14 geographical attributes. Commercial approaches for travel recommendations range from The tourism domain is a popular branch of recommender sys- merchants focusing on the sale of travel-related services, such as ac- tem (RS) research because it is a highly emotional, personal, and tivities, transport and lodging, to review platforms with a business inherently complex topic. Early systems recommended single items, model based on commissions. Depending on the type of business, such as attractions or bundled travel packages [17], and there are travel recommendations are a side-product or a main feature in big commercial players, such as hotels, restaurants, airlines, and which the recommendation can include a single product or service activities. In their survey, Borràs et al. [3] categorized an intelli- or complete trips. Big platforms, such as TripAdvisor and Google gent tourism RS into four functionalities: travel destination and Maps, recommend separate activities to users based on ratings, tourist packs, suggested attractions, trip planners, and social as- reviews, and behavior on the platform. Social networks, such as pects. In 2014, most approaches focused on the attraction suggestion Facebook, provide less structured ways to ask friends for travel category; however, currently, the trend is on complex recommenda- recommendations as a way to provide crowd-sourcing recommen- tions [30], such as sequences of attractions [29], composite travel dations. Google Trips recommends single- or multi-day tours [8] regions [6, 13], and group recommendations [5] for tourism. When in the vicinity based on user behavior and by scanning the user’s it comes to complex recommendations such as enjoyable routes, booking confirmations in Gmail. the challenge is to identify relevant points of interest (POIs) and Mafengwo4 and Qyer5 (both solely available in Mandarin) are then connect them in a coherent trip. This problem is called the the closest platforms to our approach and provide travel-related Tourist Trip Design Problem (TTDP) [9], which is algorithmically services, as well as trip planning, and sharing functionalities. interesting and has been widely investigated [28]. However, in this paper, we tackled the complexities of travel using a crowdsourcing approach by performing personalized travel 4 AUGMENTING WONDARY’S TRAVEL recommendations using actual trips from users. Crowdsourcing RECOMMENDATIONS has the advantage of being able to vary the length of travel, such As described in this paper, its core functionality is the first step as a multi-month world trip, a week trip to an island, or a weekend in wOndary’s travel RS. To answer the second part of RQ1, this in a city, and this is an unsolved challenge in the tourist RS for section discusses wOndary’s future agenda concerning trip rec- solving the TTDP. Furthermore, the structured representation of ommendations. We plan to improve the item categorization, to trips allows the combination of several independent blocks into a enable explanations and critiquing of our recommendations, and to prolonged trip or the possibility of selecting parts of a trip if the explicitly support the travel decision-making process for groups. traveler is short on time. Determining the duration of stay at each As discussed at the end of Section 2, currently, new users are location can be further personalized with additional information not provided with content-based recommendations. We believe we about the traveler, such as tourist mobility patterns [7] from past can overcome the cold start problem with an elaborate click stream trips. analysis and an initial preference elicitation phase in which users The RS of static travel items utilizes ratings as one factor of a provide their feedback for the five categories e.g., through small hybrid recommendation algorithm [4]. However, because we ex- games. ploit the trip structures to aggregate and reassemble trips, ratings The current categorization is based on expert knowledge and are not of much use due to their high sparsity. Furthermore, we data sources for categorization. It would be useful to do a thor- are concerned that users are not motivated to provide ratings for ough investigation of the attraction’s attributes with unsupervised trips and blocks, and the platform’s user experience could decline learning to obtain data-backed clusters. Furthermore, a latent factor if it required users to rate trips. Therefore, we have employed the analysis of the trips would be interesting to evaluate the explicit content-based recommendation paradigm [23] to match items to categories. As we have rich information about the trips, the core of users. Content-based recommendations are commonplace as a hy- our recommender system is content-based. This could be improved brid factor in complex domains, such as in scientific publications [1], in the future with more hybrid factors, e.g., knowledge-based rec- news articles [15, 16] or tourism [14]. However, for a purely content- ommendations and collaborative ratings of items. To provide trans- based recommendation, it is often challenging to model the user parency and improve trust in the recommendations, it would be after the very same features as the items to compute a similar- highly interesting to provide explanations of the recommenda- ity measure, e.g., the cosine distance, for ranking items. When tions [27] to the user. These explanations could be based on the investigating potential classification schemes of touristic items for classification of items ("because you liked . . . " ), the users ("travelers content-based recommendations, the work of Neidhardt et al. is an similar to you also liked . . . "), or by taking the social network on established alternative to wOndary’s categorization. Based on the the platform into account ("your friend traveled to . . . ") [2]. Another Big Five Factor Model [18] from personality psychology and prior promising technique to improve recommendations is critiquing [19]. research on tourist roles [10, 31], Neidhardt et al. developed the A conversational element [20] within the presentation of results Seven Factor Model of tourist behavioral patterns [21]. In a follow-up would enable active learning of user needs [24]. This is useful be- study [22], they showed that this can be used to elicit user pref- cause we think that it is unlikely that recommendations in such erences via pictures classified by domain experts. However, the a complex domain are perfect on the first iteration, e.g., because final step of using these tourist behavioral patterns to recommend 4 http://www.mafengwo.cn/ items was only recently performed [25] and required a very big 5 http://www.qyer.com/ RecTour 2018, October 7th, 2018, Vancouver, Canada. 16 Copyright held by the author(s). travelers may want to go on a different type of holiday than they [10] Heather Gibson and Andrew Yiannakis. 2002. Tourist Roles: Needs and the went on before. Lifecourse. Annals of Tourism Research 29, 2 (April 2002), 358–383. [11] Google Consumer Insights. 2016. Travel Micro-Moments Guide. Online. (July Since travel planning on wOndary is already collaborative, it is a 2016). https://www.thinkwithgoogle.com/marketing-resources/micro-moments/ logical step to extend the recommendations to groups to support the micro-moments-reshaping-travel-customer-journey/ [12] Daniel Herzog. 2017. Recommending a Sequence of Points of Interest to a Group decision-making process. However, we acknowledge that this issue of Users in a Mobile Context. In Proceedings of the Eleventh ACM Conference on has not been resolved and is still of high interest to the research Recommender Systems (RecSys ’17). ACM, New York, NY, USA, 402–406. https: community in this area [5, 12, 26]. //doi.org/10.1145/3109859.3109860 [13] Daniel Herzog and Wolfgang Wörndl. 2014. A Travel Recommender System for Combining Multiple Travel Regions to a Composite Trip. In CBRecSys@RecSys 5 CONCLUSIONS (CEUR Workshop Proceedings), Vol. 1245. CEUR-WS.org, Foster City, Silicon Valley, California, USA, 42–48. In this paper, we described our approach for recommending high- [14] Daniel Herzog and Wolfgang Wörndl. 2016. Extending Content-Boosted Col- quality crowdsourced trips. We presented a novel structure for laborative Filtering for Context-aware, Mobile Event Recommendations. In Pro- itineraries that is both useful for users to obtain an understanding ceedings of the 12th International Conference on Web Information Systems and Technologies. SCITEPRESS - Science and and Technology Publications, 293–303. of trip characteristics and for software systems to work with. This https://doi.org/10.5220/0005763702930303 data model structures user-submitted trips, thereby defining items [15] Philip Lenhart and Daniel Herzog. 2016. Combining Content-based and Col- laborative Filtering for Personalized Sports News Recommendations. In CBRec- based on different lengths, i.e., trips, blocks, and days. Second, we Sys@RecSys (CEUR Workshop Proceedings), Vol. 1673. CEUR-WS.org, 3–10. have automatically classified the aforementioned item types using [16] Jiahui Liu, Peter Dolan, and Elin Rønby Pedersen. 2010. Personalized News wOndary’s categorization scheme. We exploited this categorization Recommendation Based on Click Behavior. In 15th International Conference on Intelligent User Interfaces (IUI'10). ACM. https://doi.org/10.1145/1719970.1719976 scheme to perform user modeling without explicit elicitation of [17] Qi Liu, Yong Ge, Zhongmou Li, Enhong Chen, and Hui Xiong. 2011. Personalized preferences using the user’s past trips. Finally, we showcased a Travel Package Recommendation. In Proceedings of the IEEE 11th International user interface for intuitively presenting recommendations for trips Conference on Data Mining (ICDM ’11). IEEE, Vancouver, BC, Canada, 407–416. https://doi.org/10.1109/icdm.2011.118 across the globe. [18] Robert R. McCrae and Oliver P. John. 1992. An Introduction to the Five-Factor The current version is the core part of the content-based rec- Model and Its Applications. Journal of Personality 60, 2 (June 1992), 175–215. Issue 2. https://doi.org/10.1111/j.1467-6494.1992.tb00970.x ommender system and will be extended with advanced features in [19] Lorraine McGinty and James Reilly. 2011. On the Evolution of Critiquing Rec- the future. While the recommendations of this platform are more ommenders. In Recommender Systems Handbook, Francesco Ricci, Lior Rokach, personalized than ranking trips by distance to the center of the Bracha Shapira, and Paul B. Kantor (Eds.). Springer US, Boston, MA, USA, 419–453. https://doi.org/10.1007/978-0-387-85820-3_13 map, we want to confirm the perceived accuracy by utilizing an [20] Lorraine McGinty and Barry Smyth. 2006. Adaptive Selection: An Analysis automated A/B testing framework, which we will also use to con- of Critiquing and Preference-Based Feedback in Conversational Recommender tinuously measure future changes in the algorithm, of which we Systems. International Journal of Electronic Commerce 11, 2 (Dec. 2006), 35–57. https://doi.org/10.2753/jec1086-4415110202 have sketched several in our future work section. This will estab- [21] Julia Neidhardt, Rainer Schuster, Leonhard Seyfang, and Hannes Werthner. 2014. lish wOndary’s test setup for travel recommendations to provide Eliciting the Users’ Unknown Preferences. In Proceedings of the 8th ACM Confer- ence on Recommender systems (RecSys ’14). ACM, New York, NY, USA, 309–312. informed decisions regarding improvement of the product. https://doi.org/10.1145/2645710.2645767 [22] Julia Neidhardt, Leonhard Seyfang, Rainer Schuster, and Hannes Werthner. 2015. REFERENCES A picture-based approach to recommender systems. Information Technology & Tourism 15, 1 (March 2015), 49–69. https://doi.org/10.1007/s40558-014-0017-5 [1] Joeran Beel, Akiko Aizawa, Corinna Breitinger, and Bela Gipp. 2017. Mr. DLib: [23] Michael J. Pazzani and Daniel Billsus. 2007. Content-Based Recommendation Recommendations-as-a-Service (RaaS) for Academia. In ACM/IEEE Joint Confer- Systems. Springer Berlin Heidelberg, Berlin, Heidelberg, 325–341. https://doi. ence on Digital Libraries (JCDL’17). IEEE, 1–2. https://doi.org/10.1109/JCDL.2017. org/10.1007/978-3-540-72079-9_10 7991606 [24] Neil Rubens, Dain Kaplan, and Masashi Sugiyama. 2015. Active Learning in [2] Philip Bonhard and M. Angela Sasse. 2006. ’Knowing me, knowing you’ — Using Recommender Systems. In Recommender Systems Handbook, Francesco Ricci, profiles and social networking to improve recommender systems. BT Technology Lior Rokach, and Bracha Shapira (Eds.). Springer US, Boston, MA, USA, 809–846. Journal 24, 3 (July 2006), 84–98. https://doi.org/10.1007/s10550-006-0080-3 https://doi.org/10.1007/978-1-4899-7637-6_24 [3] Joan Borràs, Antonio Moreno, and Aida Valls. 2014. Intelligent tourism recom- [25] Mete Sertkan, Julia Neidhardt, and Hannes Werthner. 2018. Mapping of Tourism mender systems: A survey. Expert Systems with Applications 41, 16 (Nov. 2014), Destinations to Travel Behavioural Patterns. In Information and Communi- 7370–7389. https://doi.org/10.1016/j.eswa.2014.06.007 cation Technologies in Tourism 2018, Brigitte Stangl and Juho Pesonen (Eds.). [4] J. H. Chen, K. M. Chao, and N. Shah. 2013. Hybrid Recommendation System for Springer International Publishing, Cham, 422–434. https://doi.org/10.1007/ Tourism. In 2013 IEEE 10th International Conference on e-Business Engineering. 978-3-319-72923-7_32 IEEE, 156–161. https://doi.org/10.1109/icebe.2013.24 [26] Kadri Sylejmani, Jürgen Dorn, and Nysret Musliu. 2017. Planning the trip itinerary [5] Amra Delic, Judith Masthoff, Julia Neidhardt, and Hannes Werthner. 2018. How for tourist groups. Information Technology & Tourism 17, 3 (Sept. 2017), 275–314. to Use Social Relationships in Group Recommenders: Empirical Evidence. In 26th https://doi.org/10.1007/s40558-017-0080-9 Conference on User Modeling, Adaptation and Personalization (UMAP ’18). ACM, [27] Nava Tintarev and Judith Masthoff. 2007. A Survey of Explanations in Recom- New York, NY, USA, 121–129. https://doi.org/10.1145/3209219.3209226 mender Systems. In 2007 IEEE 23rd International Conference on Data Engineering [6] Linus W. Dietz. 2018. Data-Driven Destination Recommender Systems. In Pro- Workshop. IEEE, 801–810. https://doi.org/10.1109/ICDEW.2007.4401070 ceedings of the 26th Conference on User Modeling, Adaptation and Personalization [28] Pieter Vansteenwegen and Dirk Van Oudheusden. 2007. The Mobile Tourist (UMAP ’18). ACM, New York, NY, USA. https://doi.org/10.1145/3209219.3213591 Guide: An OR Opportunity. OR Insight 20, 3 (July 2007), 21–27. https://doi.org/ [7] Linus W. Dietz, Daniel Herzog, and Wolfgang Wörndl. 2018. Deriving Tourist 10.1057/ori.2007.17 Mobility Patterns from Check-in Data. In Proceedings of the WSDM 2018 Workshop [29] Wolfgang Wörndl, Alexander Hefele, and Daniel Herzog. 2017. Recommending a on Learning from User Interactions. Los Angeles, CA, USA. sequence of interesting places for tourist trips. Information Technology & Tourism [8] Zachary Friggstad, Sreenivas Gollapudi, Kostas Kollias, Tamas Sarlos, Chaitanya 17, 1 (Feb. 2017), 31–54. https://doi.org/10.1007/s40558-017-0076-5 Swamy, and Andrew Tomkins. 2018. Orienteering Algorithms for Generating [30] Min Xie, Laks V.S. Lakshmanan, and Peter T. Wood. 2010. Breaking out of the Travel Itineraries. In Proceedings of the Eleventh ACM International Conference on Box of Recommendations: From Items to Packages. In Proceedings of the Fourth Web Search and Data Mining (WSDM ’18). ACM, New York, NY, USA, 180–188. ACM Conference on Recommender Systems (RecSys ’10). ACM, New York, NY, USA, https://doi.org/10.1145/3159652.3159697 151–158. https://doi.org/10.1145/1864708.1864739 [9] Damianos Gavalas, Charalampos Konstantopoulos, Konstantinos Mastakas, and [31] Andrew Yiannakis and Heather Gibson. 1992. Roles tourists play. Annals of Grammati Pantziou. 2014. A survey on algorithmic approaches for solving Tourism Research 19, 2 (jan 1992), 287–303. https://doi.org/10.1016/0160-7383(92) tourist trip design problems. Journal of Heuristics 20, 3 (June 2014), 291–328. 90082-z https://doi.org/10.1007/s10732-014-9242-5 RecTour 2018, October 7th, 2018, Vancouver, Canada. 17 Copyright held by the author(s).