RDF Applications with Prolog
by Bijan Parsia | Pages: 1, 2, 3, 4, 5, 6

At this point, I can parse an RDF/XML document into a list of Prolog terms using the (now consulted) predicate load_rdf/2. I'm going to use as my example an RSS 1.0 file from XMLhack.com.

?- load_rdf('bijan/xmlhack.rss', List_of_RDF_statements).

List_of_RDF_statements = [rdf('http://xmlhack.com/rss10.php', rdf:type, 'http://purl.org/rss/1.0/channel'), rdf('http://xmlhack.com/rss10.php', 'http://purl.org/rss/1.0/':title, literal('XMLhack')), rdf('http://xmlhack.com/rss10.php', 'http://purl.org/rss/1.0/':link, literal('http://xmlhack.com/')), rdf('Description__1', 'http://purl.org/dc/elements/1.1/':resource, literal('http://xmlhack.com/')), rdf('http://xmlhack.com/rss10.php', 'http://purl.org/dc/elements/1.1/':source, 'Description__1'), rdf('http://xmlhack.com/rss10.php', 'http://purl.org/rss/1.0/':description, literal('Developer news from the XML community')), rdf('http://xmlhack.com/rss10.php', 'http://purl.org/dc/elements/1.1/':language, literal('en-us')), rdf('http://xmlhack.com/rss10.php', ... :..., literal(...)), rdf(..., ..., ...)|...]

Yes

I've not yet "consulted" the RDF file, I've merely parsed it into a list of terms. What's the difference? For one, none of these statements can be queried, e.g.,:

?- rdf(X,Y,Z).

ERROR: Undefined procedure: rdf/3

Thus far, I haven't ever (in this session) asserted any clause of an rdf/3 predicate, so the system has no special understanding of terms like rdf('http://xmlhack.com/rss10.php', rdf:type, 'http://purl.org/rss/1.0/channel). But, in the symbolic processing language tradition, that term is a well-formed data structure (indeed, it's a "term"), and it can be manipulated in various ways -- including being asserted as a clause of a predicate. In other words, Prolog has a uniform representation of code and data and facilities for manipulating a term from either perspective. Since rdf_load/3 returned a list of terms that are perfectly legal Prolog clauses, it's no problem to assert them into the current knowledge base, simply using the assert/1 predicate.

?- load_rdf('bijan/xmlhack.rss', [H|_]), assert(H).	Both the file at that relative path must be parseable into a list of terms with at least one item. The first item is bound to Hwhile the tail of the list is bound to the "anonymous", "throw away", or "don't care" variable_ and (remember a comma between predicates in a query is read as "and") that first term must be entered in our knowledge base.
H = rdf('http://xmlhack.com/rss10.php', rdf:type, 'http://purl.org/rss/1.0/channel') Yes	The query succeeded with the first parsed item from the document is bound to H.
?- rdf(X,Y,Z).	Are there any X, Y, and Z that exist in an rdf relation?
X = 'http://xmlhack.com/rss10.php' Y = rdf:type Z = 'http://purl.org/rss/1.0/channel' ; No	Yes, there's one with the listed bindings, but if I ask for more (e.g., by entering the semicolon) Prolog doesn't find any.

Two points. First, using rdf_load/3, I can transform a certain serialization of a set of RDF statements into a form that Prolog can handle well (i.e. a list of rdf/3 based terms). Second, given the appropriate output predicates, I can use that form to compose RDF as well. After all, I can assert any rdf/3 statement that I'd like at the query prompt, e.g.,

?- assert(rdf(a, b, c)).

Yes

And I can test to see if it's in fact in the knowledge base:

?- rdf(a, b c).

Yes

Since that's not a particularly useful RDF statement, I'll just retract it:

?- retract(rdf(a, b, c)).

Yes

?- rdf(a, b c).

no

I really don't want to assert these statements tediously one at a time at the query prompt. But I can get a list of the rdf/3 terms, which suggests that I could use some sort of "mapping" predicate to assert all the items in the statement list. checklist/2 takes a predicate and applies it to each member of a supplied list:

?- load_rdf('bijan/xmlhack.rss', List_of_RDF_statements),
| checklist(assert, List_of_RDF_statements).

(Note, the | is the "continuation" prompt in SWI-Prolog and not a bit of syntax.)

After this query, I have quite a few RDF statements in my Prolog knowledge base:

?- rdf(X,Y,Z).

X = 'http://xmlhack.com/rss10.php'
Y = rdf:type
Z = 'http://purl.org/rss/1.0/channel' ;

X = 'http://xmlhack.com/rss10.php'
Y = rdf:type
Z = 'http://purl.org/rss/1.0/channel' ;

X = 'http://xmlhack.com/rss10.php'
Y = rdf:type
Z = 'http://purl.org/rss/1.0/channel' ;

X = 'http://xmlhack.com/rss10.php'
Y = 'http://purl.org/rss/1.0/':title
Z = literal('XMLhack') ;

...etc.

This is equivalent to consulting a file containing all the rdf/3 items, written just as they appear in the list. Thus, one could use the rdf/3 notation to compose RDF documents (although this notation isn't all that much more convenient than the dreaded RDF/XML).

Pages: 1, 2, 3, 4, 5, 6

Next Page