Section: New Results
Keywords : Corporate Memory, Corporate Semantic Web, Cognitive Sciences, Knowledge Acquisition, Knowledge Management, Knowledge Engineering, Ontology, Assistance to the User, Cognitive Psychology, Communication, Co-operation, Human-machine interaction, Virtual Community, Community of Practice.
The objective of this research direction is to study various forms of human interoperability (e.g. search / annotation human interoperability, users' scenarios / developers' scenarios interoperability), so as to specify and to implement corresponding articulating functionalities. We will perform experimentations on different kinds of semantic distances. Moreover, we aim at extending the current interaction toolkit (SeWeSe) with new visualization techniques (e.g. statistical views on ontology-based representations) and new interaction channels (e.g. mails, IRC). We also intend to generalize ontology-based reasoning for smarter interfaces.
We are implementing a cooperative ontology editor, named ECCO, dedicated to support end-users with different profiles (domain expert, engineer, ontologist, ...) in a cooperative process of ontology construction and evolution. This editor is a new version of the previous OntologyCreator editor and it is based on:
the first users' feedbacks on OntologyCreator prototype,
the new users needs that have appeared,
a more friendly and integrated (relative to the future project's portal) user interface,
the Semtags library.
The idea of SweetWiki is to revisit the design rationale of wikis, taking into account the wealth of new standards available for the Web eleven years later to address some of the shortcomings identified through experience. SweetWiki relies on Web standards for the wiki page format (XHTML), for the macros included in pages (JSPX/XML tags), for the semantic annotations (RDFa, RDF), for the ontologies it manipulates (OWL Lite), etc. It improves access to information with faceted navigation, enhanced search tools and awareness capabilities, acquaintance networks identification, etc., [Oops!] , [Oops!] , [Oops!] .
It also provides a single WYSIWYG editor for both metadata and content editing, with assisted annotation tools (auto-completion, checkers for embedded queries or annotations). It comes with an embedded ontology editor and a reasoning engine. Finally it allows metadata to be extracted and exploited by other applications. By semantically annotating the resources of the wiki and by reifying the wiki object model itself, SweetWiki provides reasoning and querying capabilities. All the models are defined in OWL Lite schemata capturing concepts of the wikis (wiki word, wiki page, forward and backward link, author, etc.) and concepts manipulated by the users (users, folksonomy, external ontologies). These ontologies are exploited by an embedded instance of Corese allowing us to support the lifecycle of the wiki, e.g., restructure pages, to propose new functionalities, e.g., semantic search, user-profile-based monitoring and notification, and to allow for extensions, e.g., support for new medias or integration of legacy software.
In SweetWiki we have paid special attention to preserving the essence of a wiki: simplicity and social dimension. Thus SweetWiki supports all the common wiki features such as easy page linking using WikiWords, versioning, etc., but also innovates by integrating a WYSIWYG editor extended to support social tagging functionalities, embedded SPARQL queries etc., masking the OWL-based annotation implementation. Users can freely enter tags and an auto-completion mechanism suggests existing ones by issuing queries to identify existing concepts with compatible labels. Thus tagging is both easy and motivating (real time display of the number of related pages) and concepts are collected in folksonomies. Wiki pages are stored directly in XHTML or in JSPX format, embedding semantic annotations (RDFa and GRDDL) ready to be reused by other software.
Sewese: Semtags and Semservices libraries
Semtags and Semservices are two libraries that allow the use of semantic Web notions and Corese software in Web applications background. Semtags is a set of JSP tags dedicated to semantic Web. This tag library provides Web developers with tools like Corese administration tasks, ontology and annotation management tasks, and tools to send SPARQL queries. This library is written in Java language and uses the JSTL library (for JSP tags part) and Corese software (as the semantic search engine).
Semservices is a set of Web services relying on the same functionalities as Semtags but dedicated to Web services architecture. This set of services is written in Java language and currently deployed with the CXF Web service container (and Apache-Tomcat).
These two libraries are used in a European project (Palette), in an ANR RNTL project (e-WOK_HUB) and in applications developed by the Edelweiss team like CoSWEM.
Designing User-Adapted Semantic Web Applications
Our goals are:
To propose methods and models to help the user-oriented design and evaluation of Semantic Web applications; in particular, importing and adapting methods from the Human-Computer Interaction (or software ergonomics) and CSCW communities to the Semantic Web community, esp. the ontology engineering community.
To study practices of users and communities of users to inform the design of Semantic Web applications.
Collaborative heuristics to evaluate collaborative ontology editors and collaborative Semantic Web applications
"Heuristic evaluation"  is a widely-spread method for evaluating the usability of human-computer interfaces. It consists of inspecting, or walking through, a given interface, with some user's scenario or task in mind, to assess if the interface complies a set of usability heuristics (or ergonomic principles) such as "Heuristic 9: Help users recognize, diagnose, and recover from errors". This technique however is more appropriate to evaluate "tools for single users" (or individual software) than to evaluate "collective tools" (or groupware). In other words, this method relies mainly on individual heuristics; it doesn't integrate collective heuristics. To overcome this limitation, Baker, Greenberg and Gutwin  ,  , developed a preliminary set of groupware-specific heuristics (e.g., "Heuristic 7: Support people with the coordination of their actions"), based on the "mechanics of collaboration", i.e., "the small-scale actions and interactions that group members must perform in order to get a task done in a collaborative fashion".
We started an action aiming at contributing to the development of heuristics for evaluating collaborative ontology editors and, more generally, collaborative Semantic Web applications. This year, as part of the e-Wok HUB project, we used and adapted Baker et al.'s heuristics to evaluate the generic version of the collaborative ontology editor ECCO developed in Edelweiss [Oops!] . Some adaptations were inspired by users' behaviors and comments during a user experiment comparing the use of ECCO and the use of a classical text editor to elaborate the hierarchy of concepts of a small ontology [Oops!] .
The « interactive document-questionnaire » technique
During the need analysis phase of any system development cycle, it is not always possible for designers/developers and users of the to-be developed system to interact face-to-face or synchronously. It is necessary to use asynchronous or distant communication modes. As part of the e-WOK_HUB project, we developed and used the "interactive document-questionnaire" technique for asynchronously performing the needs analysis of a collaborative ontology editor. The aim of this technique is to allow users and designers/developers to get a clear and shared vision of: (a) the kind of editor to design/develop, (b) the types of users of the editor, (c) the types of ontologies that the users intend to construct, (d) the ontology building process (i.e., tasks and scenarios) that the users want to follow, and (e) the kinds of functionalities and user interfaces that could be expected by the users. The interactive document-questionnaire is iteratively filled by the users in interaction with the designers/developers who can ask users new questions to refine and clarify the points considered as important [Oops!] .
State-of-the-art of collaborative ontology editors
For the most, existing ontology editors are not collaborative, or (as Protégé and OilEd) they provide a tiny support to the collaborative development of ontologies. As part of the e-WOK_HUB project, we performed a state-of-the-art of existing ontology editors which integrate collaborative aspects, and, to complement this state-of-the-art, we also reviewed other kinds of collaborative tools such as: collaborative editors of knowledge bases, collaborative text/graphic editors, collaborative learning tools, collaborative argumentation tools, and workflow management tools [Oops!] . This state-of-the-art has been used as a need-analysis instrument. It was distributed to the e-WOK pilot users to help them specify the adaptations to be made to the generic version of the ECCO ontology editor, and to indicate the functionalities or interface components they want or do not want to see included in the editor [Oops!] . One of the critical needs identified is that the tool allows a flexible (or not strictly sequential) workflow of ontology construction steps1.
Scenario-based design of semantic services and applications
In the framework of the Palette project, we contributed to the elaboration of scenario-based techniques (definition of scenario templates, construction of scenario-based evaluation questionnaires and procedures) for designing and evaluating CoP-dedicated services, especially KM services. We applied these techniques to elaborate and evaluate the Palette CoP scenarios to be supported by the services [Oops!] . Note that these techniques are part of a larger participatory design methodology.
Survey of graph-visualization and graph-edition user interfaces
As part of the Color action Griwes, we initiated a survey of graph-visualization and graph-edition user interfaces. The aim of this survey was to help specify the "User Interfaces" level of the graph platform Griwes.
Knowledge Management Services for Communities of Practice
Keywords : Semantic Web, Ontology, Semantic Annotation, Cognitive Sciences, Assistance to the User, Cognitive Psychology, Co-operation, Human-machine interaction, Virtual Community, Community of Practice, Web service.
This work takes place in the framework of the Palette IST project, aimed at supporting communities of practice (CoPs), by offering them Knowledge Management services.
Ontology for Communities of Practice
We integrated the different subontologies related to the main concepts of Community, Actor, Competency, Resources so as to constitute the O'CoP ontology dedicated to communities of practice (see http://ns.inria.fr/palette/2007/07/ocop ). The method for the collaborative building of the O'CoP ontology, elaborated by Edelweiss, as well as O'CoP main concepts and our experience feedback for each phase of the O'CoP building process are described in the Palette deliverable [Oops!] , and synthesized in [Oops!] , [Oops!] .
The ECCO tool that supports the ontology development method, has been used by some partners as well as some CoPs mediators - for the results' validation phases -. These uses enabled us to collect the users' feedbacks and improve ECCO interface and functionalities.
Basic Knowledge Management Web services
Knowledge Management Services rely on a set of basic operations, that were implemented in the top of the Sewese (see section 5.2 ) library.
These implemented services offer the following functionalities :
Ontology management: to create, remove and modify ontologies in RDF/S and OWL. http://argentera.inria.fr/semservices/OntologyEdition?wsdl
Annotation management: to manage annotations http://argentera.inria.fr/semservices/Annotation?wsdl
Retrieval: to perform semantic search using the SPARQL Query Language, and based on the semantic search engine Corese. http://argentera.inria.fr/semservices/Query?wsdl
Administration: to factorize some administrative operation needed by all the other services. http://argentera.inria.fr/semservices/Admin?wsdl
Collaborative knowledge creation services
The tool SweetWiki has been deployed for 9 CoPs involved in the Palette project. Some preliminary results and observations on its use by some of these CoPs are synthesized in [Oops!] and described in the Palette deliverable [Oops!] .
Additional functionalities for the collaborative work and learning using SweetWiki were identified by Edelweiss in collaboration with the CoPs using it, they mainly concern:
semantic awareness functionalities: enabling users to subscribe to the set of predefined queries that meet their respective needs, so as to receive notification mails with information about the changes on the wiki content and the statistics that they are interested in.
and ease of use through the improvement of the tagging mechanisms and tag-based search, and an enhanced tag-management interface (ergonomics, drag & drop mechanism for structuring the ontology, enabled multiple inheritance, etc.).
Participant : Kalina Yacef.
This work was carried out in the framework of Kalina Yacef's visit in the Edelweiss team. Within the Palette project and its Evaluation Service, we have explored how mining the data of CoPs using SweetWiki could support the evolution of this particular Knowledge Service, so that its functionalities and quality remain high at all times and that its maintenance is manageable. As a community-owned Web service, there are needs to maintain the evolving knowledge contained in SweetWiki (in the structural form of Webs, wiki pages, tags and ontologies as well as the more implicit form of thematic networks) as it is dynamic in nature and is subject to becoming redundant and obsolete. There are also needs to monitor the evolution of the community itself and its expertise network in order to provide better user adaptation.
We have identified four broad goals of support where data mining can contribute:
Support for semantic annotation and ontology maintenance;
Support for building thematic networks within the community;
Support for building user profile and identification of community experts;
Support for the evaluation of community resources.
To support these goals, we mainly selected the following data mining techniques and visualization techniques: clustering, association rules, frequent sequential pattern mining, standard data exploration techniques (such as histograms, statistics) and specific visualizations (a timeline tree-based visualization, a social network analysis-based visualization).
The analysis using real CoP data will be conducted at the University of Sydney in December 2007-January 2008.
Participant : Isabelle Mirbel.
Researches in the field of method engineering aim at providing efficient solutions to built, improve and support the evolution of software development processes. Our background is more precisely in the domain of situational method engineering, which focuses on developing, customizing and configuring a situation-specific method from parts of existing methods.
In this research area whose main motivations are adaptability and reuse, we were more specifically interested by usage concerns.
In our previous work, we proposed a framework to combine method building and method configuration. This framework is based on a method chunk repository and a kind of ontology allowing to capitalize contextual information to refine method chunk and reuse situation while building or configuring methods.
In the continuation of this work, our current consideration focuses on means to spread and share knowledge about practices in software development. In other words, our new goal is to transform tacit knowledge about reflection on practice into explicit knowledge in order to support experience-based learning [Oops!] .
The concept of community of practice sounds very promising in this context as it refers to the process of social learning that occurs when people sharing common interest in some subject or problem collaborate over an extended period to share ideas, find solution and build innovation.
Due to the rise of the Web, several online professional communities dealing with software development have emerged. They are communities of people who organize themselves and interact primarily through the Web for work and knowledge sharing. Open source software communities are typical examples of such communities.
These communities generate huge amount of information as result of their interactions. This information is mostly structured in order to be quickly reused (mailing lists, forums, etc.). There is few means, (like FAQ for instance) to capitalize the information over a longer period of time and to turn it into knowledge (through a semantic FAQ for instance). Moreover, the capitalized knowledge mainly deals with the discussion topics (i.e., the system under development) and few about the members of the community.
In this context, our current goal is to take advantage of the semantic Web and the social Web concepts and technologies to improve the support provided to these communities in terms of long-term knowledge capitalization about the resources and the members of the community.
We plan to work on means to make members expertise, commitments and relationships among them explicit. One of our aim will be for instance to allow members to better understand who knows what in the community in order to improve coordination among members. Another one will be to explicit the expertises present in the community to let the members better understand how to contribute and to encourage them to do it.
We also plan to work on means to summarize and organize the knowledge in order to improve long-term capitalization support. One of our goal will be for instance to provide summary to discussion threads in forum as well as alternative means to organize and present discussion threads.
Ontology-driven Dynamic Course Generation for Web-based Education
Actually, the Web is becoming a de-facto standard platform for providing various kinds of educational resources to support teaching in a university or a technical training company. One of the greatest benefits of the Web is that the course material created to support a specific course no longer remains the only educational material that the students can use during the course. Thus, all these resources can be reused and shared in an e-learning context. However, it is very difficult for learners and even for formation responsible to identify the resource relevance to share and reuse it. In order to reduce this problem complexity, we work just with the Web repositories like ARIADNE, MERLOT, etc. which already offer a certain organization of learning resources to facilitate their selection and access.
In general, the learning resources, stored in Web repositories, are first subject to a pedagogy engineering work in order to give them reusable in the context of a particular formation. This activity is time and effort-consuming. In our work, we propose a different approach that consists of moving this engineering effort from the formation responsible/ expert to the software system. It consists of delivering adaptive courses directly and with a minimum human effort. This leads us to have recourse to semantic Web technologies.
We aim to make the formation process most efficient by using the semantic structure of the domain ontology and the resource annotations. So, we offer to the learner adaptive learning paths according to his level of knowledge, goal and time constraints. The learner is assisted to construct a correct representation of the domain knowledge. This representation, called an ACM (Adaptive Cognitive Map), is first calculated by applying filters/rules to the domain ontology and next revised by analyzing, in semi-automatically manner, the learner paths. This analysis can lead to the ontology or/and the resource annotations evolution.
Concretely, we have developed an adaptive learning organizer, called « Organisateur de parcours adaptatifs de formation» (OrPAF). It allows automatic generation of an individualized cognitive map taking into account a specific learning goal, the level of the learner's knowledge and his temporal constraints. Adaptive learning resources are searched from Web repositories like ARIADNE or created locally by our experts and are attached to the generated ACM. The idea of an ACM is to generate dynamically, for each learner, both an individualized course structure and course content by selecting the most optimal learning topics/concepts (knowledge) and materials (presentation, example, test or problem) at any moment. The optimal learning concept and material are selected to bring the learner closest to the ultimate learning goal. This approach is well suited for individual students taking a self-study distance-learning course. These students can be employees in an organization who have different experience and background knowledge, or students in an online university with different backgrounds and goals.