Presented at the Semantics Conference September 2021 as part of DBpedia Day. A presentation by Margaret Warren of ImageSnippets on how ImageSnippets works with DBpedia.
1. Margaret Warren
Metadata Authoring Systems
Institute for Human & Machine Cognition
Anchoring Images
to Meaning
Using DBpedia
Image by Travis S. CC BY-NC Flickr via ImageSnippets
2. Margaret
Warren
Artist/Technologist
Creator of ImageSnippets
(https://imagesnippets.com)
Research Associate,
Institute for Human and
Machine Cognition
Research started circa 2004 building concept maps around images.
Our work has been to apply formal semantic theory to the informal
ways people describe images and use RDF techniques to build image
graphs. Since the beginning of our journey, DBpedia has been a
primary source of the entities in our linked data descriptions.
4. What is ImageSnippets?
• Grew out of research around formalizing image descriptions circa 2004
• Was developed as an experimental platform 2010-2013 for untrained
users to build structured semantic annotations for images in the RDF
syntax --- we train users to construct annotations in subject-predicate-
object constructions
• Has grown into a mature framework for research on semantic annotation,
ontology engineering and knowledge graph construction with images.
Image metadata is stored in openly published datasets as well as with each
image as JSON-LD and RDFa and embedded in the images.
• Was expanded in 2020 to allow for Dockerized ML Classifiers to be added
and used to augment image annotation and triple formulations.
• Open for non-commercial research, non-profit and personal use.
(https://www.imagesnippets.com)
Bounding Ambiguity from SAD workshop 2018 HCOMP http://ceur-ws.org/Vol-2276/paper5.pdf