FollowLinksForMoreInformation is the general practice of performing web retrieval on URIs in a knowledge base to obtain more knowledge. It might also be called "Inclusion During Query Answering" and is sometimes loosely called FollowYourNose. It gives rise to SelfAnsweringQuestion's and the SelfDescribingWeb. From a multi-agent system design angle, this sometimes looks like PullTheDefinition.
There are many specific practices one might follow here, such as only following links in the RDF predicate position, or only following rdfs:seeAlso links; see LinkProperties for discussion. The myriad options have different results in two dimensions:
- Performance. Following all the links may take too long or
produce too much data, so the question is which links to follow first or follow at all. From this angle, links indicate relevance, pointing out some information which (true or false) is probably more interesting than other information of unknown veracity.
- Trust. What to do with any knowledge you
gather? Do you simply consider it true since UseImpliesConsent (There are degrees of consent and trust. Trust is not boolean. guest), do you consider it provisionally true in some sense, do you consider some of it true based on what kinds of links were followed, ...?
Just use the complete WebClosure, including everything, trusting everything.
The DefinitionalWebClosure is more restricted, but is currently untested.
DocRulesClosure is another experimental approach.
Closer to applications, we can imagine this might manifest as:
All SemanticWeb applications might issues WARNINGS about any logical inconsistencies found among the closure of nearby documents.
Query anwering servers might offer answers coming from the deductive closure of nearby documents, but qualify such answers as second-rate; users/clients could ask for details in the proof (as with any other answer).