Between May and August 2014, I worked on a Google Summer of Code project to incorporate metadata from the Commons into the structured content collected by DBpedia. I worked on a Github fork, using Github issues to track the project. This work has now been published in the 14th International Semantic Web Conference, Bethlehem, PA, USA (doi:10.1007/978-3-319-25010-6_17).

Deliverables

Completed

Added support for the Wikimedia Commons to the DBpedia Extraction Framework
- Modified the Extraction Engine to allow for the first-ever generation of template statistics on the Commons: #2, #4
- Activated extractors that already produce meaningful data from Commons pages: #5, #17, #23

Added support for files to the DBpedia Extraction Framework
- Created a FileTypeExtractor that extracts filenames, file extensions and MIME types as triples: #3, #11
- Extended the ontology for describing media classes used in the Commons, including: dbo:File, dbo:Image, dbo:StillImage, dbo:MovingImage and dbo:Sound: #10

Designed and built new extractors
- Created a GalleryExtractor to extract galleries from pages and store them using the dbo:hasGalleryItem property: #26
- Created an ImageAnnotationExtractor that stores annotations made using the ImageAnnotator gadget using the W3 Media Fragments recommendation: #31
- Created a CommonsKMLExtractor that extracts overlay.kml files stored on the Commons to provide overlays to Commons images: #22

Made it easier to create and test template mappings to Commons templates
- Added support for Commons templates that include other templates, such as {{Self}}: #16
- Improved the extractionSamples page on the mappings server to support any namespace, not just Main, for example: extractionSamples in the File namespace for {{Gray's Anatomy plate}}: #9
- Created a generic identifier property (dbo:identifiedBy) and a category for identifiers on the Mappings wiki: #14
- Incorporated and extended my mentor's code for adding prefixes and suffixes to a property to support ObjectProperties: #29

Mapped some Commons templates to test the above, including:
- Automatically mapped around 360 license templates: #18, #20
- {{VN}}: #12

Examined the file metadata dump to see if there was any interesting metadata there (there wasn't): #21
Handle disambiguation pages on the Commons (there are only ~5,000, so we decided not to work on this): #24
Extract image captions from the language Wikipedias (ran out of time): #27
Writing tests for the FileTypeExtractor and LabelExtractor: #8, #25
Mapped template mappings for the top-10 most used templates (ran out of time): #7
Propose a new scheme for linking objects in DBpedia through URI-based identifiers to the rest of the Web of Things (ran out of time): #15

A test dataset of 2,033 files I used to test my code.
The RDF output produced from this test dataset.
A list of every major license type used in the Commons, how it can be identified, and how it can be represented in DBpedia: #34