In this paper we illustrate Tourpedia, which would be the DBpedia of tourism. Tourpedia contains more than half a million places, divided in four categories: accommodations, restaurants, points of interests and attractions. They are related to eight locations: Amsterdam, Barcelona, Berlin, Dubai, London, Paris, Rome and Tuscany, but new locations are continuously added. Information about places were extracted from four social media: Facebook, Foursquare, GooglePlaces and Booking and were integrated in order to build a unique catalogue. Tourpedia provides also a Web API and a SPARQL endpoint to access data.
Tuscany, Italy. For more details about datasets about tourism, please refer to: http://datahub.io/dataset?q=tourism.
In this paper we illustrate Tourpedia, which would be the DBpedia of Tourism. Tourpedia is reachable through its portal6 and is available also in the datahub.io platform7.
Tourpedia was developed within the OpeNER Project8 (Open Polarity Enhanced Name Entity Recognition), whose main objective is to implement a pipeline to process natural language.
The usage of Tourpedia could be very various. For example, it could be used to perform named entity disambiguation in tourism domain, or to extract the most appreciated points of interest in a town. 2
6 http://tour-pedia.org 7 http://datahub.io/dataset/tourpedia 8 http://www.opener-project.eu 9 http://www.facebook.com 10 http://foursquare.com 11 https://plus.google.com/u/0/local 12 http://www.booking.com 13 http://tour-pedia.org/about/statistics.html 14 http://tour-pedia.org/gui/demo/
Social Media
Data Extraction OpeNER pipeline
InteDgartaation Reviews
Places NE Repository Linked Dataset SPARQL endpoint D2R server
Web API
Web Application
The sentiment of a place is calculated as a function of all the sentiments of the reviews about that place. In order to retrieve the sentiment of a review, the OpeNER pipeline was used. In particular, each place is associated to zero or more reviews extracted from social media (i.e. Facebook, Foursquare and Google Places). Each review is processed through the OpeNER pipeline and is associated to a rate, which expresses its speci c sentiment. 2.2
Linked Data
Tourpedia is exposed as a linked data node and provides a SPARQL endpoint15.
The service is implemented through the use of a D2R server16. For each place, the
following ontologies are used to represent it: VCARD [
Web API Tourpedia provides a RESTful API18 to access places and statistics. The output of each request can be JSON, CSV and XML. For example, a search request about Places is an HTTP URL of the following form:
where parameters must be at least one one of the following: location (the location of the places), category (the type of the places such as accomodation), attraction, restaurant, poi), and name (the keyword to be searched). 15 http://tour-pedia.org/sparql 16 http://d2rq.org/ 17 http://wiki.dbpedia.org/Ontology 18 http://tour-pedia.org/api In this paper we have illustrated Tourpedia, which would be the DBpedia of Tourism. It could be interesting a deeper connection between Tourpedia and DBpedia. At the moment, in fact, only locations are connected to DBpedia. As future work, we are going to align also attractions and points of interest contained in Tourpedia to DBpedia.
Tourpedia could be exploited both by tourism stackholders to get the sentiment about touristic places and by common users.
At the moment, the procedure to update datasets is manual. As future work, we are going to de ne a semi-automatic procedure to update them and to add new locations.
This work has been carried out within OpeNER project, co-funded by the European Commission under the FP7 (7th Framework Programs Grant Agreement n. 296451).