Linked-AI 2010, Stanford, March 2010
The Problem: Choose a Vocabulary
Assume you want to choose a vocabulary (microformat, microdata, RDFa, standards) to mark up some structured data in HTML (or even some other format).
A user wants to use RDFa to mark up something as simple as some personal information attached to a review, how should they do it?
- FOAF: http://xmlns.com/foaf/spec/
- W3C RDF mappings of vCard: http://www.w3.org/TR/vcard-rdf
- Google's vocabulary: http://rdf.data-vocabulary.org
Too Many URIs!
There are too many URIs for the same thing...
An average of 1,339 URIs (S.D. 8,000) returned per query. That's a lot, but most may obviously be URIs that just mention the term...
Top Entities and Concepts
Number of Searches (RDF Hits) Entity:
- 7311 (99) david blaine
- 2997 (134) jessica alba
- 2100 (16723) nick
- 1280 (178) michael hayden
- 1098 (10) marcus vick
- 1092 (199) keith urban
- 1015 (43) lane bryant
- 990 (55) desmond dekker
- 922 (312) jennifer white
- 900 (100) clay aiken
- 883 (359) bill cosby
Concepts:
- 11383 (10767) weather
- 10321 (7777) dictionary
- 3675 (434333) people
- 3217 (189115) music
- 3117 (7196) monster
- 2192 (1444) autism
- 1468 (149436) map
- 1198 (17562) travel
- 1191 (12067) pregnancy
- 1104 (82074) news
The power-law blues
Very few empirical studies have been done on Linked Data, and often of the type "Look, it's a power-law!" without any proper statistical tests. See Clauset, Newman et al. paper referenced in my paper for a better method!
Entity Queries: alpha = 2.31, with long tail behavior starting around a frequency of 17 and a Kolmogorov-Smirnov D-statistic of .0241, indicating a significant good fit.
Concept Queries: The alpha= of the queries for concept queries was calculated to be 2.12, with long tail behavior starting around a frequency of 36 with a Kolmogorov-Smirnov D-statistic of .017.
Hypertext Query Frequency and Linked Data?
Question: Is the amount of Linked Data returned correlated with the popularity of the query?
No: Spearman's rank correlation statistic was the insignificant .0077 (p > .05), while for concept queries, the correlation was the still insignificant at .0125 (p > .05)
Hypertext Query Frequency and Linked Data?
Question: Is the amount of Linked Data returned correlated with the popularity of the query?
No: Spearman's rank correlation statistic was the insignificant .0077 (p > .05), while for concept queries, the correlation was the still insignificant at .0125 (p > .05)
Vocabulary Hosting Requirements
- Use of URIs : Each term in a vocabulary has a URI.
- Persistent URIs: Each term URI will be used to refer to the same term in perpetuity and will not be repurposed.
- Persistent documentation : Each term URI should remain resolvable to `namespace documents' -- descriptive documentation in HTML and/or machine-readable representations such as RDF schemas.
- Change policy : The stability of the meaning of the terms should be determinable -- i.e., the meanings of terms should evolve according to known change management policies and with responsibility for changes traceable to either individuals or organizations.
- Analytics : Vocabularies should be plugged into actual empirical data about their usage in the wild, allowing vocabulary producers and consumers to see how the vocabulary is actually used. This is invaluable for future versioning of the vocabulary and for discovering valuable patterns in the vocabulary usage.
Vocabulary Preservation
How can multiple vocabulary hosts be decentralized?
- Contracts : Provide a well-understood framework for contracts between Vocabulary Hosting Services and the Vocabulary Preservation System, defining processes and governance for handling rapid interventions (e.g., redirecting URIs if disaster strikes) and for transferring ownership and maintenance responsibility in the long-term.
- Open : Offer itself as a partner to any Vocabulary Hosting Service that tries to aim for best practices in vocabulary hosting, and so serve as a further `seal-of-approval' for Vocabulary Hosting Services.
- Reliable: Aim at ensuring that at least all large-scale (popular) vocabularies have such mechanisms in place.
- Meta-review : Provide regular reports on the status of maintenance activities for covered vocabularies, possibly as a group with other Vocabulary Preservation Systems. As a by-product of this reporting, the VPS or coalition of VPS systems would function as a clearinghouse for information about vocabularies.
Linked Data Cloud as of 2007
Linked Data Cloud as of Now-ish (late 2009)
The Modest Proposal
- Domain-name : A neutral domain name should be used -- one that belongs to a non-profit organization or could be transferred to one. Suggestions have been sharednames.org, semanticweb.org, a new domain name, or another domain name of an existing body.
- Legal Body : Some sort of legal body needs to be host the domain name or purchase. Another option could be an existing organization, and the W3C or IANA would be possibilities.
- Technical Infrastructure: The site would need a strong policy on service and maintenance, with a robust and extensible infrastructure. Organizations like IANA or the W3C do not have at the moment the technical resources to maintain such a vocabulary hosting service by itself.
- Long-term financial stability: Since managing persistent URIs and hosting documentation takes resources, there needs to be some baseline financial model. This can initially be supported by work from interested parties, and if enough interested parties sign up, the domain name can be guaranteed for a long period of time.
- Clear Social Process: There are two main options: In the open model, any person could easily submit, update, and comment upon a vocabulary, with a simple line of responsibility going to a single individual or the VHS itself. In the working group model, a group of interested parties is chaired for some period of time to come to consensus on a vocabulary.
Who will do this?
Any takers?