Free text search on Musicbrainz literals using Virtuoso RDF Views

I introduced a Virtuoso RDF View that maps the Musicbrainz relational database into RDF using the Music Ontology a couple of weeks ago. Now I will show some query examples evolving a special feature of these Virtuoso RDF Views: full text search on literals.

How RDF Views work

A Virtuoso RDF View can be seen as a layer between a relational database schemas and its conceptualization in RDF. The role of this layer is to convert relation data in its RDF conceptualization.

That is it. You can see it as a conversion tool or as a sort of lens to see RDF data out of relation data.

How full text search over literals works

Recently OpenLink Software introduced the full text feature of their Virtuoso’s SPARQL processor with the usage of the “bif:contains” operator (it is introduced into the SPARQL syntax like a FILTER).

When a user sends a SPARQL query using the bif:contains operator against a Virtuoso triple store, the parser will use the triple store’s full text index to perform the full text search over the queried literal.

With Virtuoso RDF View, instead of using the triple store’s full text index, it will use the relational database’s full text index (if the relational database is supporting full text indexes, naturally).

Some queries examples

In this section I will show you how the full text feature of the Virtuoso RDF Views can be used to increase the performance of a query against the Musicbrainz RDF View modeled using the Music Ontology

Note: if the system asks you for a login and a password to see the page, use the login name “demo” and the password “demo” to see the results of these SPARQL queries.

Example #1

A user remember that first name of the music artist is Paul, and he remember that one of the albums composed by this artists is Press Play. So this user wants to get the full name of this artist with the following SPARQL query:

sparql
define input:storage virtrdf:MBZROOT
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX mo: <http://purl.org/ontology/mo/>
PREFIX dc: <http://purl.org/dc/elements/1.1/>
SELECT ?artist_name ?album_title
FROM <http://musicbrainz.org/>
WHERE
{
?artist rdf:type mo:SoloMusicArtist .
?artist foaf:name ?artist_name .
?artist mo:creatorOf ?album .

?album rdf:type mo:Record .
?album dc:title ?album_title .

FILTER bif:contains(?artist_name, “Paul”) .
FILTER bif:contains(?album_title, “Press and Play”) .
};

Results of this query against the musicbrainz virtuoso rdf view

As you can notice with that query, the user will use the full text capabilities of Virtuoso over two different literals: the objects of these two properties foaf:name and dc:title.

Example #2

In this example, the user wants to know the name of the albums published by Madonna between 1990 and 2000. The answer to this question is returned by the following SPARQL query:

sparql
define input:storage virtrdf:MBZROOT
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX mo: <http://purl.org/ontology/mo/>
PREFIX dcterms: <http://purl.org/dc/terms/>
prefix dc: <http://purl.org/dc/elements/1.1/>
SELECT DISTINCT ?albums_titles ?creation_date
FROM <http://musicbrainz.org/>
WHERE
{
?madonna rdf:type mo:SoloMusicArtist .
?madonna foaf:name ?madonna_name .
FILTER bif:contains(?madonna_name, “Madonna”) .

?madonna mo:creatorOf ?albums .
?albums rdf:type mo:Record .
?albums dcterms:created ?creation_date .
FILTER ( xsd:dateTime(?creation_date) > “1990-01-01T00:00:00Z”^^xsd:dateTime ) .
FILTER ( xsd:dateTime(?creation_date) < “2000-01-01T00:00:00Z”^^xsd:dateTime ) .
?albums dc:title ?albums_titles .
};

Results of this query against the musicbrainz virtuoso rdf view

Here the user will use the full text capabilities of the Virtuoso RDF Views to find artists with the name Madonna and he uses two filters on xsd:dateTime objects to find the albums that have been created between 1990 and 2000.

Examples #3

In this last example, the user wants to know the name of the members of the music group U2.

sparql
define input:storage virtrdf:MBZROOT
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX mo: <http://purl.org/ontology/mo/>
SELECT ?band_name ?member_name
FROM <http://musicbrainz.org/>
WHERE
{
?band rdf:type mo:MusicGroup .
?band foaf:name ?band_name .
?band_name bif:contains ‘”U2″‘ .
?band foaf:member ?members .
?members rdf:type mo:SoloMusicArtist .
?members foaf:name ?member_name .
};

Results of this query against the musicbrainz virtuoso rdf view

Here the user will use the full text feature to get the name of the music group, then the name of the members related to this (these) music group(s) will be returned as well.

Special operators of a full text search

Some full texts operators can be used in the literal parameter of the bif:contains clause. The operators are the same used in the full text feature of Virtuoso’s relational database. A list and a description of the operators can be found on that page.

I would only add that the near operator is defined as +/- 100 chars from the searched literal. And the wildcard ‘*’ operator should at least be placed after the third character of the literal. So, “tes*t” or “tes*” or “test*” are legal usages of the wildcard operator, but “*test”, “t*” or “te*st” are illegal usages of the operator.

Conclusion

Finally, as you can see, the full text feature available with the Virtuoso RDF Views is a more than essential feature that people should use to increase the performance of their SPARQL queries. The only two other options they have are: (1) using a normal “literal” that as to be well written and with the good cases; in one word this option render such queries useless and (2) they can use a FILTER with a regular expression with the “I” parameter that is far too slow for normal usages.

6 Responses to “Free text search on Musicbrainz literals using Virtuoso RDF Views”


  1. 1 Gunnar Grimnes May 17th, 2007 at 5:51 pm

    There is also a implementation of full-text search using lucene on top of the Sesame RDF store, if you’d like another RDF DB :)

    http://gnowsis.opendfki.de/wiki/LuceneSail

    We are currently working on rewriting it for working with the new Sesame2Beta version.

  2. 2 Mike Bergman May 17th, 2007 at 7:41 pm

    Fred,

    Excellent post! These examples are very helpful to people like me that are still learning SPARQL and trying to get used to the whole RDF triple thing.

    I think Gunnar’s comment is interesting as well. It is clear these triple stores are becoming serious data management tools with full-text indexing and the works. Of course, Virtuoso is virtually untouchable in regards to the numbers of formats and data models that it supports!

  3. 3 Victor Aug 23rd, 2007 at 11:54 am

    Great post. this is gonna be helpful for places like peekamo. P.S I totally agree with hthe fact that virtuoso is a flawless model support system.

  4. 4 Neha Mar 25th, 2014 at 2:52 pm

    Hi Fred,

    This was really helpful to me. I am new to RDF and exploring ways to so full text search. We are using OpenRDF and sesame workbench. I want to know if there exists any virtuoso plugin for any of these. I will be really helpful to me. I could not find much on the web about it.
    Thanks

  5. 5 Neha Mar 25th, 2014 at 4:44 pm

    Hi,

    Can you also throw light on “How can we do set up of virtuoso with sesame”. It will be really helpful to me.

    Thanks

  6. 6 Frederick Giasson Jun 4th, 2014 at 9:53 am

    Hi Neha,

    I unfortunately never configured Sesame with Virtuoso. However, I know that many people did, so some research on Google should return the relevant information.

    Good luck!

Leave a Reply






This blog is a regularly updated collection of my thoughts, tips, tricks and ideas about my semantic Web researches and related software development.


RSS Twitter LinkedIN


Follow

Get every new post on this blog delivered to your Inbox.

Join 69 other followers:

Or subscribe to the RSS feed by clicking on the counter:




RSS Twitter LinkedIN