• LarkcProject
• WP6
• Communication
• 2010-01-20_Munich-meeting

WP6 face-to-face meeting, Munich, January 20th 2010

Participants:

• CEFRIEL: Emanuele, Irene
• Saltlux: Kono, Stanley
• Siemens: Florian, Werner

Agenda:

• selling points for the review

• AMA vs. ?OpenStreetMap/LinkedGeoData

• integrating traffic prediction within LarKC
• social recommendation
• D6.7 - 3rd report on data and performances (M26 - May 2010)
• next steps
• AOB

Discussion minutes:

• selling point for the review (and for us internally as well)
  • last year: using LarKC to realize a use case scenario in Urban Computing
  • big claim: making it easier to
    • develop data-intensive mash-ups
    • on massive, heterogeneous, evolving (on which we want to predict future states or say something on the present state), possibly inconsistent and incomplete data
    • with LarKC rather than with other approaches
    • reasons: software engineering perspective; RDF, SPARQL, linked data; data scale, heterogeneity, evolving data, mixing static and dynamic data, mixing static and probabilistic data; semantically-rich data; exploiting free/open data (which can be interesting for companies), mixing proprietary and open data
  • goal for the review:
    • porting the alpha urban LarKC in Korea
    • traffic data for Milano
• exploration of OSM/LGD by Saltlux and comparison with AMA data
  • LGD has 2B triples; way and nodes primitives, which ontologically match with (and link to) DBpedia model
  • comparison: work needed to get which "nodes" in the 2 models are the same; precision in coordinates is different so experiments are needed to match the 2 datasets
  • mapped to Google Maps to overlay the points from the 2 datasets to "see" the mismatches

  • ACTION ITEM (Manu): look for streets in OSM and to check the ?TerraCognita paper (about OSM --> program to convert to KML --> conversion to LGD)

  • ACTION ITEM (Irene/Daniele): check API to see if there is any way to extract streets
    • done: current APIs have no method to extract streets; however, original OSM source has easy-to-translate XML for streets
  • ACTION ITEM (Kono/Stanley): try the following SPARQL query to exploit the link between LGD and DBpedia: show me the shape and give me the information of the attraction in Milano called "Duomo" in DBpedia; more complex query: show me the shape and give me the information of an attraction in Milano which has a capacity greater than 20.000 persons
  • collaboration with COREONTO project (run by KAIST in Korea): semi-automatic ontology instantiation
    • ACTION ITEM (Kono/Stanley): check if they can help in extracting events from zero.eu website (or similar sources)
• traffic prediction by Siemens
  • re-exporting of data from the map: besides the data already present before, we now have streets and part-of-street relationship, traffic sensors, link-type
  • moreover, there are time series of sensors data (conceptually linked to sensors on the map)
  • proposal for new traffic prediction pipeline and related issues/problems
    • split/join within pipeline to handle our cases

    • batch calculation of traffic prediction (instead of live query) --> scheduled execution of plugins

    • smart way to model and transfer/translate data: what is the best possible way? trade-off between expressiveness and response time
• D6.7 - 3rd report on data and performances (M26 - May 2010)
  • ACTION ITEM (Irene/Emanuele): ask Frank/Alice to postpone the delivery from May to July (M28), in order to being able to make evaluation with the new platform to be released in March
    • done
  • ACTION ITEM (Irene/Daniele): ask Alexey about the work done by HLRS on parallelization and if we can compare performances before/after parallelization on their machines (since it is not useful to compare it with performances on our servers...)
    • done, Matthias Assel is responsible for this; we will keep in touch for further development/testing
  • ACTION ITEM (Kono/Stanley): if not in a previous deliverable, describe OSM/LGD data
  • ACTION ITEM (Florian): if not in a previous deliverable, describe traffic/weather data
  • ACTION ITEM (Florian): check if we can add a new pipeline that plays with the traffic data (probably too early for traffic prediction pipeline)
  • ACTION ITEM (Florian/Ralph): strategy to evaluate traffic-aware path finding and find a reference to traffic-aware path finding