The tripscore Linked Data client: calculating specific summaries over large time series

David Chaves Fraga, Julian Rojas, Pieter-Jan Vandenberghe, Pieter Colpaert, Oscar Corcho

Jul 28, 2017 (modified: Aug 01, 2017) ISWC 2017 DeSemWeb Submission readers: everyone
  • Abstract: Time series – such as public transport time schedules and their actual departure times – may deliver insights about the public transport network to third parties. Today, however, public transport data is published in a way in which analytical processing is too expensive. In previous work, the Linked Connections framework was introduced as a cost-efficient publishing alternative to the de-facto GTFS standard and route planning APIs. We study whether this server interface can also be used by Linked Data agents to solve analytical queries over longer periods of time. In this work, we created a serverless Linked Data client in Javascript for the analysis of time series on top of public transport data sources, called tripscore.eu. In this example, it calculates the quality of experience for your journey for the last 5 weeks using the public transport agencies it can discover. We have made the code to this proof of concept available as open source in different reusable components. As the user-perceived performance is quite slow, we formulate opportunities to achieve better response times. We could, on the one hand, suggest the data publisher to publish summaries over longer periods of time. On the other hand, we could also, as reusers, create a private summary of the data on our server and expose this to our user agents. Still an open issue is how this client would discover new public transport agencies reliably, for which we started working on a metadata profile for transport datasets.
  • TL;DR: Calculating specific summaries over large time series on transport domain with Linked Connections approach
  • Submission category: Intelligent Client Challenge / Demo
  • Url: https://htmlpreview.github.io/?https://github.com/dachafra/tripscore/blob/master/DeSemWeb17.html
  • Authorids: dchaves@fi.upm.es, julianandres.rojasmelendez@ugent.be, pieterjan.vandenberghe@ugent.be, pieter.colpaert@ugent.be, ocorcho@fi.upm.es
  • Keywords: Linked Data, time series, decentralization, route planning

Loading