Section: New Results

Scalable and Expressive Techniques for the Semantic Web

The team has continued developing expressive models and scalable algorithms for exploiting Semantic Web data, in particular RDF graphs, as well as rich corpora consisting of Web documents with semantic annotations.

We have studied efficient algorithms for answering RDF queries in the presence of schema (or semantic) constraints such as described through the RDF Schema language. The difficulty here consists of efficiently taking into account the data that is implicitly present in the RDF database due to semantic constraints, and which needs to be reflected in query results. We have identified the expressive database fragment of RDF, which extends previously identified fragments of the RDF specification by allowing more expressive schema and queries, and provided novel efficient algorithms for answering Basic Graph Pattern queries (a popular dialect of the standard SPARQL query language) over RDF graphs pertaining to the RDF Database Fragment. Our query answering algorithms take advantage of the processing power of a relational database management system while also reflecting RDF semantics [25] .

The ability to exploit large corpora of heterogeneous RDF data requires tools for analyzing RDF content through the lenses of a specific user perspective, or user need. Such tools are commonplace in the context of relational data management, where data warehousing is a well-developed area, but lack completely in the realm of RDF. We have proposed a novel framework for building and exploiting all-RDF data warehouses [33] and have implemented this framework in a proof-of-concept platform [32] . A main contribution of this work is to preserve RDF graph structure, heterogeneity, and rich semantics from the base data to the analytical schema and analytical schema instance. Thus, our proposal is the first to allow the analysis of rich Semantic Web (RDF) data while preserving its rich content and semantics. For more information on this project, see https://team.inria.fr/oak/warg/ .

We have investigated the usage of semantics as a way to enrich, interconnect, and interpret rich corpora of Web data. In particular, within the XR project, we had proposed in prior work the XR (XML+RDF) data model which integrates XML documents and RDF triples treating both as first-class citizens. One particular use of XR is to annotate nodes in XML documents, by RDF triples which may for instance describe their properties or state how nodes are semantically related to some concept or to each other. In [18] we describe the data model and core query language, make a comprehensive analysis of query evaluation algorithms, and describe extensive experiments carried within a fully implemented platform, as part of the PhD thesis of J. Leblay [12] . The XR platform was put to task in an application context related to digital journalism, where an XR content warehouse is continuously enriched through document analysis and annotation. This scenario has lead to a software demonstration [24] , [35] and a keynote tutorial [38] . In collaboration with A. Deutsch, we have extended the XR query language and provided query-view composition algorithms in [41] .