diveplusdata

repository for datasets for the DIVE Plus demonstrator See http://diveproject.beeldengeluid.nl. Currently we have partial data from four collections and three vocabularies. These are in RDF Turtle.

Open Images

Consisting of the video collection itself as well as the GTAA thesaurus Open Images collection Originally from openbeelden.nl (3000 videos) Creative Commons – Attribution-Share Alike license (CC-by-SA) Subset in DIVE triple store is all videos from OpenImages, as found in an existing RDF conversion. For 510 videos crowd enrichments are added with relations entities, events, persons, places through crowdsourcing.

GTAA (Gemeenschappelijke Thesaurus Audiovisuele Archieven). From http://gtaa.beeldengeluid.nl . Entities from KB are also matched to this thesaurus.

All Open images content can be found in the oi_dive folder:

• al_oipers_to_gtaapers.ttl alignments from crowdsourced persons to GTAA
• al_oiplace_to_gtaageo.ttl alignments from crowdsourced locations to GTAA
• al_oirest_to_gtaaall.ttl alignments from crowdsourced concepts to GTAA
• gtaa_clean.ttl.gz GTAA dump
• oi_collectiontriples.ttl triples registering media objects to a collection
• oi_enrichments.ttl crowdsourced enrichments
• oi_schema_mappings.ttl OI to Dive schema mappings
• oi_sources.ttl for sources and placeholders (thumbnails and video locations)
• openbeelden.ttl Openimages collection, earlier dump by Michiel Hildebrand
• void.ttl VOID file, includes library statements for ClioPatria

Radio bulletins from KB

Originally from http://www.delpher.nl/nl/radiobulletins. Size is 1.5 millions of items ANP has made the objects (JPGs, OCRs, ALTOs) in this set available under a CC-BY-NC-ND 3.0 license (link is external).

These are all news bulletins between 1937 (start of collection) and May 1955 (when my script crashed initially).

• The original XML docs were retrieved from the KB server using OAI-PMH (my application crashed when this arrived at May '55).
• The XML was converted to RDF using a python script: https://github.com/biktorrr/dive/blob/master/kb/get_kb_data_new.py This script actually calls another KB API which does NER on the OCR'ed text of the bulletin Each bulletin becomes a MediaObject and links to exactly one Event (the news event mentioned in the bulletin'. The label of this Event is a 'best guess' based on the OCR'ed description of the bulletin. This of course leads to many 'hard to read' labels, but that is something we need to learn to live with. I improved the labeling a bit. An example of such a label is http://purl.org/collections/nl/dive/kb/evt-anp:1953:11:19:56 or http://purl.org/collections/nl/dive/kb/evt-anp:1953:02:15:49 Persons, places and actors found in the text are saved as Dive entities and linked to the Event. For example, the location 'Middelburg' http://purl.org/collections/nl/dive/kb-loc-Middelburg ** Persons, places and actors are aligned with GTAA
• This results in 197,199 Media Objects (around 80 more than previous collection) 5,389,556 RDF triples in the main KB graph (now the largest named graph in the triple store: http://data.dive.beeldengeluid.nl/browse/list_graphs ** Around 6,000 alignments

OLD: Subset in DIVE triple store is 2210 digitized typoscripts (radio news scripts, to be read during news broadcasts) from the period 1937-1984.These were chosen to match the Openbeelden subset by re-using terms found there for the search request.

Amsterdam museum

Originally from https://bitbucket.org/biktorrr/amlod (thousands of objects) Which was derived from public collection website (www.amsterdammuseum.nl/collectie/) Data published under Creative Commons Attribution license (CC-by)

The collection and vocabularies are found in the am_dive folder.

Collection Subset Now, all objects from the datadump are in. In total, there are 73,447 objects with 5.7Million triples.

Amsterdam Museum Thesaurus the thesaurus used to annotate AM objects. This is (partially) aligned with GTAA.

Tropenmuseum

Originally from the old 2004 data we converted for the eculture project (eculture.multimedian.nl). This consists of some 70.000 objects. This is most likely the same as http://www.opencultuurdata.nl/wiki/tropenmuseum/. That dataset has CC-by-SA 3.0.

Changes made are a) added new locations of images based on the updated wereldculturen.nl api. image URIs are based on work IDs and b) changed the namespace of kit triples from hash to slash URIs and removed whitespaces from URIs

The collection and vocabularies are found in the kit_dive folder.

Collection subset Now all objects are in the dataset. This adds up to 78,270 objects and 1.9 Million triples

SVCN the thesaurus used to annotate Tropenmuseum objects. This is (partially) aligned with GTAA. There is also a list of actors (for now unaligned)