BIS PhD Seminar – Processing and Evaluating Large Amounts of Open Data – My Progress

View Github Repository Open presentation in a new window

earthquakesan

See all presentation from earthquakesan

BIS PhD Seminar – Processing and Evaluating Large Amounts of Open Data – My Progress

0 0

PhDSeminar-03032014

Presentation for PhD Seminar on 3rd March 2014

On Github earthquakesan / PhDSeminar-03032014

BIS PhD Seminar

Processing and Evaluating Large Amounts of Open Data

Created by Ivan Ermilov 3rd March 2014

Who Am I? What is hidden behind the topic name?

Retrospective

  • Last PhD Seminar - 3rd December 2012
  • AKSW Colloquium
  • I am a part of AKSW/BIS group - since September 2011
  • What's the progress so far?

My Progress

  • Projects
  • Publications
  • Other Activities

Projects

  • SlideWiki
  • CubeViz
  • LODStats (current)
  • CSV2RDF Mass Conversion Framework (current)

Publications

  • SlideWiki: Elicitation and Sharing of Corporate Knowledge using Presentations by Ali Khalili, Sören Auer, Darya Tarasowa, and Ivan Ermilov in Proceedings of the EKAW 2012
  • Towards an Open-Governmental Data Web by Ivan Ermilov, Claus Stadler, Michael Martin, and Sören Auer in Proceedings of the KESW2012
  • Linked Open Data Statistics: Collection and Exploitation by Ivan Ermilov, Michael Martin, Jens Lehmann, and Sören Auer in Proceedings of the 4th Conference on Knowledge Engineering and Semantic Web
  • CSV2RDF: User-Driven CSV to RDF Mass Conversion Framework by Ivan Ermilov, Sören Auer, and Claus Stadler in Proceedings of the ISEM '13, September 04 - 06 2013, Graz, Austria

Other Activities

Participated in the Semantic Web Challenge 2013 with the CSV2RDF project.

Planned and pending publications

  • A Fine-Grained Evaluation of SPARQL Endpoint Federation Systems by Muhammad Saleem, Yasar Khan, Ali Hasnain, Ivan Ermilov, Axel-Cyrille Ngonga Ngomo (submitted to Semantic Web Journal)
  • CubeViz journal publication (Michael Martin)
  • LODStats journal publication

Work in progress

LODStats project: a statement-stream-based approach for gathering comprehensive statistics about RDF datasets.

CSV2RDF project: user-driven semantic mapping of tabular data.

LODStats

  • Evaluate RDF datasets triple by triple
  • Gathers 32 stat. criteria such as:
    • Number of triples, entities, literals
    • Average string length
    • Vocabularties, classes used
  • Helps to understand the data
  • Generates VoID desctiptions

LODStats Architecture

LODStats Features

  • Report of warnings and errors for datasets
  • Report on statistical criteria for datasets
  • Export as VoID/DataCube metadata
  • General LOD cloud statistics
  • Search function fro datasets, vocabularies, classes, properties, languages, datatypes
  • REST interface for the search functions
  • Linked Data publication of statistics
  • SPARQL endpoint
  • CubeViz installation for the statistics visualization

LODStats Hidden Features

  • Which visualization is suitable for my dataset?
    • RDF DataCube Vocabulary/Ontology used?
    • Contains geospatial information?
    • What is the class hierarchy depth?
  • Does dataset fit into the specific application?
    • Linked to DBPedia?
    • Largest known dataset in the domain?

LODStats: Future Steps

  • Parallel processing
    • Inside the dataset
    • Distributing datasets across the cluster
  • LinDA project (visualization framework)

CSV2RDF Mass Conversion Framework

377 data catalogs are available with a large amount of tabular data. How to deal with that data? (Only 3% are available as RDF)

Results so far

  • CSV resources converted: 15627
  • CSV resources volume: 62 GB
  • Number of generated triples: 7.3 B

Architecture

Generated Mappings

User Interface

Open Questions

  • Schema/ontology matching
  • Classification of the tables
  • Instance matching
  • Interface to combine these three steps

Questions?

Presented BY Ivan Ermilov, BIS/AKSW