1 Introduction

Although the Semantic Web is meant to help machines interpret data on the Web, end-users still need to view and query the data. Ideally, also non-experts should be empowered to formulate queries. State of the art retrieval systems for semistructured data such as [1] that are targeted towards end-users still use keyword-based search interfaces. We believe that there is a need for graphical interfaces that enable non-expert users to formulate complex queries over semistructured data.

Specifying a graphical notation for a query language involves a trade-off between user interface complexity and language expressiveness. At one end of the spectrum are systems such as Magnet [2] and multi-faceted browsing and navigation systems such as Flamenco [3]. In both system one uses so-called facets as restrictions to filter the data set. These systems are limited in that they do not allow object nesting; only filters pertaining to one item can be specified. In addition they need domain specific customizations tailored towards the schema definition of the data set.

On the other end of the spectrum is Query-by-Example (QbE) [4] known from databases. QbE is a complete query language for the relational calculus, can express a wide range of relational queries, and includes insert and update operations. There is no need to customize that language to a specific domain, since the queries are constructed based on the database schema. However, since semistructured data typically comes without a fixed schema, it is not possible to directly apply the QbE paradigm here.

The goal of this poster is to present a graphical notation for queries over semistructured data that identifies a middle ground between the two paradigms which we believe is useful and easy to use -- and provides sufficient expressiveness to cover a wide range of queries.

We introduce our graphical notation by means of examples. The exemplary data set is expressed in RDF (Resource Description Framework) using Dublin Core and FOAF vocabularies.

2 Preliminaries

In the following, we introduce the basic ideas for the representation. We assume that the reader has rudimentary knowledge of RDF, a language to represent semistructured data on the Web.

The atomic unit of RDF are RDF triples, which consist of subject, predicate, and object. We use the Notation3 (N3) syntax that introduces variables to the RDF data model to be able to specify queries. The basic notion of an RDF query is an RDF triple pattern. A triple pattern is a triple where subject, predicate, or object can be a variable. An N3 query consists of a ql:where clause with one or more triple patterns specifying the selection criteria, and a ql:select clause which specifies the format of a query result.

In the following, we define the notion of an RDF facet which is the main element of a graphically constructed query.

Definition 1 (RDF Facet)  Given a set of URI references , a set of literals , and a set of variables , a triple (s, p, o) is called an RDF facet.

A facet can be seen as a filter condition over an RDF graph. Multiple facets can be combined on subject and object positions by using the same variable, which amounts to a join.

3 Walkthrough

In this section, we introduce the different building blocks of our graphical notation by means of examples. We begin each example with a textual description of the query, then show the corresponding graphical representation, and finally present the query expressed in N3 query syntax. In the queries we omit namespace declarations for brevity.

Single Facet.

<> ql:select {
  ?subject1 ?p ?o .
}; ql:where {
  ?subject1 foaf:name "Andreas Harth" .
  ?subject1 ?p ?o .
} .

<> ql:select {
  ?subject1 ?p ?o .
}; ql:where {
  ?subject1 dc:title ?keyword .
  ?keyword yars:keyword "computer" .
  ?subject1 ?p ?o .
} .

<> ql:select {
  ?subject1 ?p ?o .
}; ql:where {
  ?subject1 dc:title ?keyword .
  ?keyword yars:keyword "RDF" .
  ?subject1 dc:subject "Query formulation" .
  ?subject1 ?p ?o .
} .

<> ql:select {
  ?object1 ?p ?o .
}; ql:where {
  ?subject1 foaf:name "Andreas Harth" .
  ?subject1 foaf:knows ?object1 .
  ?object1 ?p ?o .
} .

<> ql:select {
  ?object1 ?p ?o .
}; ql:where {
  ?subject1 foaf:name ``Andreas Harth'' .
  ?subject1 foaf:knows ?object1 .
  ?object1 foaf:interest ?keyword .
  ?keyword yars:keyword ``DERI'' .
  ?object1 ?p ?o .
} .