A More Decentralized Vision for Linked Data (2018)
An paper giving an overview of the "Semantic Web" dream, reasons why it failed and possible solutions.
Title indicates that solutions might be found in more decentralization. The paper does however not make any concrete suggestions towards more decentralization. I find the title slightly misleading.
The two main suggestions are:
- Take inspiration from Gene Ontology (GO)
- Use HDT
The analysis of problems with currently published Linked Data is valuable:
Challenges
Technical
Availability
Size and Scalability
They talk about the challenges when dealing with billions of triples.
Not yet a problem for us.
They do propose using HDT.
Findability and (Meta-)Data formats
Datasets are hard to find/discover.
They mention VoID and Semantic Sitemaps as possible solutions.
For openEngiadina the social aspect might help the discoverability. You discover content via your peers.
“RDF Data Quality” of Datasets and the “Semantics of Links”
The usual problems with "dirty" data. In the Semantic Web world dirty data is even worse as it breaks the whole point of the Semantic web.
References Weaving the Pendantic Web.
"No clear guideline with regards to ownership of namespaces"
Interesting notes on "recipes" for namespaces.
Usually one uses
#
or/
to separate identifiers from prefixes.But some ontologies do weird things:
http://bioonto.de/sbml.owl#Uniprot:Q9UJX6
. It is unclear what is namespace prefix and what is the identifier.
Non-Technical
Most interestingly: Trust.
Developers and publishers need to trust the technology but also the people hosting the namespaces, ontologies used in your published data.
This seems solvable technically by decentralizing ontologies. It also solves the economic problem of who hosts the ontologies: People publishing data using an ontology.