ONTOSEARCH2 fact-file

What's the Problem?

Ontologies are important components of web-based applications. While the Web makes an increasing number of ontologies widely available for applications, retrieving knowledge from ontologies in the Web becomes a more challenging issue. Existing approaches are mainly based on keyword based searches of ontology metadata. Furthermore, querying of ontologies requires them to be downloaded into a DL reasoner, so that their entailments can be calculated. Using a variety of novel approaches, ONTOSEARCH2 allows a large and growing repository of ontologies from the semantic web to be rapidly queried.

Towards a Solution

ONTOSEARCH2 uses semantic approximations of description logics to reduce the complexity of ontologies. The resulting approximations are in the form of DL-Lite ontologies [Calvanese et. al] which can be stored in a relational database and can be queried with PTIME complexity (compared with exponential complexity in the case of unmodified OWL-DL ontologies).
This technique makes it practical to store and query many thousands of ontologies on a single server. Query answering over this repository is performed using SPARQL queries which are mapped onto one or more conjunctive queries that are compatible with DL-Lite query expansion algorithms. Each query is expanded and the resulting set of SQL queries is executed, and the union of results passed back to the client (either as HTML or XML).

Description of System

ONTOSEARCH2 is written as a set of Java servlets and JSP files. There are two primary interfaces to the query tool: a web based interface which uses HTML forms and returns results as an HTML table, and a Web Service interface which uses the SPARQL protocol to return results to applications wishing to integrate with the ONTOSEARCH2 system.

The ONTOSEARCH2 HTML interface
The other main function provided by the ONTOSEARCH2 site is to add new ontologies to the repository. This is done promptly, so that queries can be run against these ontologies without waiting for an auxiliary process to download and process the file offline.

Try

ONTOSEARCH2 is currently available for public beta testing at http://www.ontosearch.org. Although the current version is a beta product, the central algorithms are stable; further additions will be to add to its usability & to further enhance performance.

Submitting a URI

You can add any Semantic Web Document which contains RDF/XML based OWL data to the ONTOSEARCH2 repository. This is done by clicking the Submit link in the menu, and entering the URI of the file or files you wish to add. All files are added on request, and you will be given feedback on progress via your browser. Please be patient with this process and do not close your browser or navigate away from this page until the page has completely finished loading.

Querying ONTOSEARCH2

All queries passed to ONTOSEARCH2 are run against the entire repository, so you can query all the information which has been added to the system. Queries are made in SPARQL language with several restrictions:

• Queries must be "SELECT ... WHERE ..." queries
• Each clause in the WHERE section must follow the following rules:
• The subject can be either a variable (e.g. ?professor) or a URI (e.g. http://xmlns.com/foaf/0.1/#Person or foaf:Person)
• The predicate must be a URI
• The object can either be a variable, a literal or a URI, except when the predicate of the query is rdf:type.

Results are returned to the user as an HTML table which contains the variable names as specified in the query, and the corresponding matching values.
An example query which can be run in ONTOSEARCH2 is this query from the Lehigh University Benchmark, which requests a list of professors at a university and their contact details:
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX ub: <http://www.lehigh.edu/~zhp2/2004/0401/univ-bench.owl#>

SELECT ?professor ?name ?email ?phone WHERE {
?professor a ub:Professor;
           ub:worksFor <http://www.Department0.University0.edu>;
           ub:name ?name ;
           ub:emailAddress ?email ;
           ub:telephone ?phone .
}

Future Work

The next release of ONTOSEARCH2 will contain a keyword based search mechanism, to allow broad searches of the entire repository for matching concepts or instances. This will use an ontology metadata database to provide additional search terms which will be combined with the original query terms and executed as a query across the repository. We believe that by combining traditional web search techniques with an ontology query engine, we will be able to improve the state of the art for semantic web search.
Additional improvements are planned for the Web Service API of ONTOSEARCH2, including a function to upload ontologies to the repository.