Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: Scientific Foundations

Keywords : Artificial Intelligence, Cognitive Sciences, Exchanges, Extraction, Knowledge Acquisition, Knowledge Management, Knowledge Engineering, Knowledge representation, Knowledge Server, Corporate Memory, Ontology, Assistance to the User, Ergonomics, Co-operation, Collaboration, Interaction Design, Conceptual Graph, Graphs, Documents, XML, RDF, OWL, Semantic Web, Annotation, Web, Web architectures, Intranet, Corporate Web, Information Retrieval, Languages, Reasoning, Scenarios, User interfaces, Distributed Services, Virtual community, Community of practice.


Knowledge Management (KM) is one of the key progress factors in organizations. It aims at capturing explicit and tacit knowledge of an organization, in order to facilitate its access, sharing out and reuse [80] . The considered organization can be an actual enterprise or a public organization, but it may also just consist of a given department or service; it can also be a group, or a community, or a virtual enterprise (made of members possibly stemming from different companies, but sharing a common interest).

The Acacia approach relied on the analogy between the resources of a corporate memory and the resources of the Web. We considered that a corporate memory can be materialized in a corporate semantic Web, that consists of [80] , [81] :

According to [106] , communities of practice (CoPs) are “groups of people who share a concern, a set of problems, or a passion about a topic, and who deepen their knowledge and expertise in this area by interacting on an ongoing basis”. CoPs can be found within businesses, across business units or across company boundaries [107] , still they differ from business or functional units, from teams and networks: people belong to CoPs at the same time as they belong to other organizational structures. An effective organization comprises a constellation of interconnected CoPs, as these are privileged nodes for the exchange and interpretation of information. CoPs preserve the tacit aspects of knowledge that formal systems cannot capture. CoPs can be considered as a means by which knowledge is “owned” in practice. Indeed, such groups allow the functions of creation, accumulation and diffusion of knowledge in organizations.

The Edelweiss project-team extends this hypothesis to virtual communities and considers that a support to knowledge management and cooperative work in a community can also rely on a community semantic Web.

The underlying research topics are:

Semantic Web application development platforms

CoGITaNT [87] is a C++ library for representing and manipulating conceptual graphs (CGs) and related objects such as graph support and rules. It offers CG operations, an inference engine and a graph editor. CoGITaNT was developed first at LIRMM in Montpellier and then in Angers University.

Protégé ( ) is a free, open source ontology editor and knowledge-base framework.The Protégé platform supports two ways of modelling ontologies with frames and OWL editors. Ontologies can be exported into several formats such as RDF(S), OWL, and XML Schema. Protégé is developed at Stanford and by a large community of developers.

Jena ( ) is a Java framework for building Semantic Web applications. It provides an environment for RDF, RDFS and OWL, SPARQL and includes a rule-based inference engine. It offers persistency through database connectors. Jena is developed by HP labs at Bristol.

Most of the current contributions on semantic Web servers focus on low-level infrastructural problems and do not really address the new interaction means provided by the semantics added by the semantic Web.

KAON [104] is an open-source ontology management infrastructure targeted for business applications. It includes tools allowing ontology creation and management and provides a framework for building ontology-based applications. Its successor, KAON2, advertises a number of features: an API for programmatic management of OWL-DL, SWRL, and F-Logic ontologies; a stand-alone server providing access to ontologies in a distributed manner using RMI; an inference engine for answering conjunctive queries (expressed using SPARQL syntax); a DIG interface, allowing access from tools such as Protégé; a module for extracting ontology instances from relational databases. Thus KAON2 server capabilities are at much lower level than the one we consider here and focus on APIs and reasoning services.

RDF Gateway is a platform for the development and deployment of Semantic Web applications and InferEd is its associated authoring environment to navigate and edit RDF documents. RDF Gateway is both a Web server and a RDF database server to support the management of metadata on the Web. RDF Gateway is also at a lower infrastructural level i.e. it is closer to the back-end and does not address user interaction and ergonomics.

Likewise, Joseki is a SPARQL Server for Jena: an HTTP and SOAP engine that supports the SPARQL Protocol and the SPARQL RDF Query language. Closer to our concern is the semantic Web application development framework ODESeW applying the Model-View-Controller design pattern to semantic Web applications.

Finally, SIMILE is a joint project conducted by the W3C, MIT Libraries, and MIT CSAIL that seeks to enhance inter-operability among digital assets, schemata, metadata, and services. SIMILE includes in particular three tools relevant to our topic here: Longwell, Semantic Bank and Piggy Bank. Longwell relies on the RDF data model to propose faceted browsing and enable users to visualize and browse RDF dataset and to build a Web site out of these data. A Semantic Bank ensures remote persistence on a server to share data and lets you publish your information, both in RDF and regular Web pages through Longwell. Piggy Bank is an extension to Firefox letting users make use of existing information on the Web and in particular to merge information from several sites and recreate a unifying customized presentation. Thus the focus in SIMILE is on data integration and not so much on using the semantics to design new interfaces.

Semantic Annotations

As detailed in [102] , annotation tools can be compared according to:

In the current state of the art, we distinguish the following approaches:

Interaction Design

Initially concerned with formal and technical aspects, the Semantic Web community recently acknowledged the necessity to take seriously into account uses and users of Semantic Web applications so that such applications can be accepted by users and their organizations. An indicator of this new concern is the emergence of scientific events such as the International Workshop on Interaction Design and the Semantic Web (2004), and its successors, the Workshops on End-user Semantic Web Interaction (2005, 2006, 2007). The aim of these workshops is to help Semantic Web application designers bring the power of the semantic Web to end-users, applying Interaction Design and more specifically Social Interaction Design. Interaction Design is the discipline of defining and creating the human interaction with digital, environmental or organizational systems. Interaction design defines the behaviors or interactions of an object or system over time with its users' population. Interaction designers create systems that are typically informed by research on users and their practices. Social interaction design accounts for interactions among users as well as between users and their devices. Social interaction design is practice-oriented. It is concerned with sign and symbolic value, social behaviors, etiquette and norms, groups and communities, structured interactions, and routines, sequencing, and temporal organization.

Interaction design is critical to a number of applications: an application may use state of the art algorithms; if it does not provide a usable interface, it will not be effective. For interactions to be supported efficiently in a community, supporting tools have to be designed taking into account the nature, the rules, the protocols, the context, etc. of these interactions. In particular, community-supporting tools must:

Interaction design and other human-in-the-loop approaches to semantic Web design

The Edelweiss project clearly fits the Interaction Design for the Semantic Web or End User Semantic Web Interaction trend, the goal of which is to optimize user interfaces for semantic Web applications, and user interactions with and through these applications. Interaction issues are sometimes expressed in terms of usability (see, e.g. [94] ). Usable Ontology is thus a method and a set of software tools, aimed at supporting domain experts in populating a domain ontology and obtaining a shared consensus on its content.

More generally the Edelweiss project fits what can be called the human-in-the-loop approach to semantic Web design, the goal of which is to adapt semantic Web applications to users and communities of users. This approach encompasses the Interaction Design trend as well as other trends such as Human Semantic Web, Pragmatic Web [79] , Sociosemantic Web [108] , and Web 2.0 or Participatory Web [99] , or Social Web.

The challenge of the human-in-the-loop approach is to effectively bring the power of the semantic Web to end-users. It requires interdisciplinarity and collaborations between Semantic Web researchers and Interaction/UI/HCI/CSCW researchers.

User interfaces to the Semantic Web

Interface languages

To hide the complexity of formal languages of the Semantic Web to the users, it has been proposed to provide the users with more understandable languages such as: (1) Natural language: see, e.g., AquaLog Q/A system or SemSearch engine; (2) Semi-structured languages: query interfaces which do not restrict the user with a formalistic language, but also impose some structure on the user’s input so as to be also domain-independent. (3) Graphical and Graph languages: In the Human Semantic Web concept browser Conzilla, RDF is combined with the more human-readable UML. Conceptual Graphs are sometimes considered as an interface language, but some authors contest its intuitiveness and propose alternative languages, like the Object-Process Methodology (OPM)-based dual graphic-textual representation, in Visual Semantic Web. In Edelweiss, we will not work directly on natural language, but on semi-structured, graphical and graph languages.

Visualizing ontologies

In this special case of the interface language issue, the goal is to make ontologies not only understandable to users, but also graspable when they are complex or large (see, e.g., the CropCircles graph visualization technique). The goal is also to make the ontology pervasive or transparent; that is, to shift end user's goal from visualizing ontologies or concepts per se (rather an ontologist's goal), to visualizing familiar objects referred to internally by ontologies or concepts. This last issue will be a major concern of Edelweiss.

Interface algorithms

Interface algorithms are those algorithms which allow, for example, to materialize users' viewpoints (e.g. view-hierarchies and multi-facet search paradigm), to support the user's query construction process (interpretation algorithms of user moves as query-construction steps), or to exploit the commonalities of the (search patterns, search terms, etc.) history of the current user with other users of the system. Collaborative filtering techniques are based on such commonalities.

Interface ontologies and ontology-based methods for designing user interfaces

Ontologies are now used to represent the world of interfaces (i.e., interface objects), or to ground methods for designing interfaces. The OntoWeaver approach, which provides high level support for Web site design and development, represents this use of ontologies very well.

Interface scenario-making and prototyping

Scenario-making [69] , [75] and prototyping, in particular paper prototyping, are widely used methods in HCI design. They are often related to participatory design. Such methods begin to be used seriously in Semantic Web interface design. For example, to better understand user expectations in annotation activities and to inform the design of annotation mechanisms.


Logo Inria