Metreeca/Self will search and navigate generic RDF data from any SPARQL 1.1 endpoint, but user experience can be improved by sticking to the following guidelines.

Annotate Content

Provide Labels

Provide human-readable labels for every graph entity by annotating them as:

<uri> rdfs:label "label" .
<uri> rdfs:label "label"@en .

Metreeca/Self will scan for available generic and english-tagged labels, in that order.

If no labels are provided, Metreeca/Self will try to infer human-readable labels for URIs according to well-known slash/hash patterns (e.g. 'property' for http://example.com/schema#property).

Provide Comments

Where relevant, provide human-readable descriptions for graph entities by annotating them as:

<uri> rdfs:comment "comment" .
<uri> rdfs:comment "comment"@en .

Metreeca/Self will scan for available generic and english-tagged notes, in that order.

Provide Pictures

Where relevant, provide visual depictions for graph entities by annotating them as:

<uri> db:thumbnail <picture-uri>/"picture-uri".
<uri> schema:image <picture-uri>/"picture-uri".    
<uri> foaf:logo <picture-uri>/"picture-uri".
<uri> foaf:depiction <picture-uri>/"picture-uri".

Metreeca/Self will scan for available pictures in this order.

Provide Coordinates

Where relevant, provide geographic coordinates for graph entities by annotating them using the Basic Geo (WGS84 lat/long) Vocabulary as:

<uri> geo:lat "latitude"; geo:long "longitude" .

Include Ontologies

Upload relevant ontologies to the endpoint: Metreeca/Self will use human-readable labels and notes for classes and predicates to configure fields and facets names and linked data pages.

Assist Navigation

Identify Key Collections

When opening a new report, Metreeca/Self will present the user with a springboard page, listing collections identified by introspecting rdf:type info from the endpoint.

To fine-tune what's presented to the user or to improve performance on large datasets, you may explicitly identify RDF classes as key collections for your dataset by annotating them as:

<{dataset}> <http://rdfs.org/ns/void#rootResource> <{class}> .

where {dataset} is the identifier of any dataset stored in the endpoint.

As far as Metreeca/Self is concerned the dataset identifier is immaterial and any class marked as a root resource will be included in the springboard page: if no more sensible option is available, just use the canonical base URL of the endpoint, e.g.:

<http://data.example.com/> void:rootResource <http://data.example.com/terms#Products> .

Include Inverse Links

Metreeca/Self supports the creation of property paths involving inverse steps (^<predicate>), but they may be very confusing even for experienced users: wherever feasible, explicitly document bidirectional relationships between graph entities using pairs of inverse predicates.

Avoid Blank Nodes

Until a consistent and efficient way to skolemize SPARQL results sets is found, blank nodes should be avoided, as Metreeca/Self won't be able to navigate past them.

Wherever feasible, replace blank nodes with named resources either identified by dereferenceable http: URIs or tagged within a suitable URI namespace (e.g. RFC 4151 - The 'tag' URI Scheme, RFC 4122 - A Universally Unique IDentifier (UUID) URN Namespace).

In any case, provide human-readable labels summarizing the information hidden beyond blank nodes or embedded within structured intermediate resources.