Project : acacia
Section: New Results
Keywords : Corporate Memory, Cognitive Sciences, Knowledge Acquisition, Knowledge Management, Knowledge Engineering, Ontology, Assistance to the User, Cognitive Psychology, Communication, Co-operation.
Support to Modelling and Building of a Corporate Memory
The objective of this action is to propose methodological and software support for the construction of a corporate memory, thanks to a user-centered approach. We study in particular the construction of a corporate semantic Web and the construction of ontologies and annotations from human and textual sources of expertise or from databases. Moreover, we study how to handle multiple viewpoimts or multiple ontologies and how to take into account the life cycle and the evolution of a corporate semantic web. We also started to study a new scenario of knowledge management: e-Learning, and to study how to use ``Semantic Web Services'' and to tackle the context and privacy of the (possibly mobile) user.
Methodology for Construction of a Corporate Semantic Web
Keywords : Knowledge Capitalization, Knowledge Management, Corporate Memory, Project Memory.
Participant : Rose Dieng-Kuntz.
We synthesized our vision of a corporate semantic web, and we proposed a life cycle based on the following steps :
Inventory of fixtures,
Choice of the application scenarios,
Building of the ontology (from human sources, form textual corpus or from structured databases),
Validation of the ontologies: consistency checking from system viewpoint, knowledge audit and validation by experts, evaluation by end-users),
Constitution, organisation and validation of resources: creation of new resources, or adaptation (transformation, reorganization) of legacy resources,
Annotation of resources: manual annotation, automatic annotation or semi-automatic annotation,
Validation of the semantic annotations and of the corporate semantic Web,
Maintenance and Dynamic evolution of the corporate semantic Web.
A synthesis of existing work on corporate semantic Webs (history, approach, methods, tools and examples of applications) is described in . Several examples of construction of such corporate semantic Webs inspired of Acacia applications are described in .
Adapting Models from Human and Social Sciences to the Design of Organizational Memory Systems
Participant : Alain Giboin.
If machine-to-machine interoperability through ontologies is a major issue of the Semantic Web research, users' mutual understanding is a critical issue. Our research actions deal with the mutual understanding issue, considered as threefold: (1) mutual understanding between users (in our case, of Semantic Web applications supporting Organizational Memory and Knowledge Management), (2) mutual understanding between users and machines, and (3) mutual understanding between users and designers. The aim of the research actions is to provide designers and users with models, methods, and tools to facilitate such types of mutual understanding.
The Correspondences framework. We further developed our work on the use of the Correspondences Framework  to analyze and support the coordination between producing and understanding or using the electronic documents which constitute a part of the memory of some organization . This dialogical framework relies on the principle that documents are props of a dialog between writers and readers of these documents, and that the dialog succeeds when correspondences can be established, through documents, between the writers and readers respective representations and processes, be they internal or external. We analyzed some organizational memory techniques in terms of correspondence processes.
The notion of Common Frame of Reference. Correspondence processes allow people to build shared internal representations which  called « [operative] common frames of reference » (référentiels (opératifs) communs), and which are also referred to as « shared context »  or « internal context » . We reviewed a set of French-speaking work in ergonomic psychology on common reference frames, as well as nonFrench-speaking work often used in this French-speaking work , and identified convergence points in this work: (a) on the nature of common frames of reference, (b) on the processes used to construct the frames, and (c) on the ergonomic consequences of (a) and (b) for designing collective work systems such as organizational memory systems or knowledge management systems. One of the convergence points we identified is the necessity to consider common frames of reference as two-sided representations, or as joint internal (mental) and external (physical) representations.
The notion of corresponding representations. Viewed as external representations, common frames of reference refer to notions such as "mediating representation", "intermediary representation", "intermediary object","boundary object", and "communication genre", which are employed in system design communities such as knowledge and ontology engineering, requirements analysis, human-computer interaction, and computer-supported cooperative work, to improve mutual understanding. For example, the knowledge engineering community uses the notion of mediating representations to denote « problem modeling languages that help bridge the gap between experts and computer implementations » . Gathering these various notions under the term of "corresponding representations", we began to analyze them using our Correspondences Framework, and relying on our previous work on representations for mutual understanding and mutual agreement . As suggested in , we also began to work towards constituting "referential components" libraries to be later on connected to "user-interface components" libraries. Scenarios are such referential components.
Scenario-based design and evaluation of ontologies and semantic web applications. This action is a follow-up to the work reported in . For the KMP project, we elaborated a scenario-based procedure to evaluate ontology-based interfaces. We also initiated a research action aimed at designing an editor of multimedia scenarios for ontology-based applications.
Generation of Semantic Annotations for an Experiment Memory: APplication to Transcriptome Analysis
Keywords : Corporate memory, Semantic Web, Natural Language Processing, Biochip experiments, Ontologies, Semantic annotations.
This work is carried out in the context of Khaled Khelif's thesis.
The study of gene expression has been greatly facilitated by biochip technology. Biochips can assess tens of thousands of genes simultaneously and provide a huge amount of information: for example, information about the roles played by particular genes in drug sensitivity and the effects of drugs on gene expression. In the framework of a collaboration project with biologists working on biochip experiments at IPMC , our work consists of assisting them in validating and interpreting the results of their experiments of obtained results. Our aim is to propose methodological and software support for the capitalization and the valorisation of knowledge resulting from experiments (semantic annotations, ontology...) and techniques to preserve and reuse data (structured documents, semantic information retrieval). We rely on the techniques of semantic web and knowledge engineering.
Initially we focused on the validation and interpretation of experiments results. Since this phase of validation is based on information retrieval, our approach rests on the semi-automatic generation of semantic annotations for scientific articles in the biochip domain. These articles can come from internal sources such as specific documentation databases for each biologist or from external sources such as on line documentation databases. Thus, we developed a system which, starting from a text written by a biologist (e.g. a scientific article), allows to generate a structured semantic annotation, based on a domain ontology, and describing the semantic contents of this text. This system called MeatAnnot, extracts information from text, instantiates concepts and relationships of UMLS ontology and generates RDF annotations for the document . Then, we validated these annotations with biologists and we tested their coherence using Corese. Finally, we are studying the scalability of the system and we are working on methods which allow to make reasoning on these annotations using the query language and rules of Corese.
Ontologies and Semantic Relation Acquisition from Biomedical Corpora
Keywords : Syntax-semantics interface, ontology and annotation learning, lexico-syntactic information extraction, semantic schema acquisition, NLP, Unification-grammars, causation, Semantic Web.
This work is performed in the framework of Laurent Alamarguy's PhD thesis. Corpus-based knowledge acquisition is an important challenge for community memory constitution by collecting and apprehending language constructions underlying domain knowledge. This constitutes a helpful method for domain experts to supervise their discovery of relevant knowledge interrelations and favors information retrieval. This work deals with the acquisition of semantic relations from biomedical corpora for the construction of ontology and annotations. This aims at elaborating methodological supports and tools to enhance the automation of ontology construction and enrichment from linguistic comprehension of texts in order to develop community memory in biomedical area.
Text comprehension through Natural Language Processing methods provides knowledge structures closer to target domain thanks to fine-grained and more accurate linguistic analysis. While respecting the domain ontological conformity, the automation of knowledge extraction and construction has to be increased by optimizing the linguistic processing, so our study naturally focuses on the syntax-semantics interface development.
After having underlined the importance of the notion of causality interrelating central nervous system pathologies and some involved genes, from our training corpus constituted with Medline abstracts, we collected some semantic markers denoting causal relations, such as verbs like "to cause", "to affect", "to inhibit", etc., or nouns like "inhibition", "triggering", etc., or more idiomatic lexical constructions like "to act upon", "to be responsible for". This step aims at discovering the nature of conceptual knowledge that could be extracted and to elaborate some conceptual frameworks by apprehending the contextual frame through the causative construction peculiar to each semantic marker with its causing and caused elements and its domain setting.
More generally our conceptual acquisition method aims at proposing some salient semantic relation schemas to domain experts who supervise the ontology and annotation construction. This method follows the linguistics processing philosophy and is declined into several main stages dedicated to the syntax-semantics interface. This interface is elaborated through the formalism of the grammatical parser, PATR-II, based on unification grammars that are performed thanks to lexicons that are constituted with the results of term extraction and shallow processing that reveals different kinds of abstract lexico-syntactic information, and thanks to grammars that we manually determine upon causative construction of salient semantic markers . This interface generates some candidate conceptual schemas that are transcribed in semantic web RDF(S) languages to be used with Corese search engine.
More particularly the focus on the syntax-semantics interface concerns text comprehension, and discourse processing is tackled through a cognitive-functional approach. We rely on this approach to elaborate the discourse parsing following the modeling of lexicon and rules taking into account some lexical and morpho-syntactic information, and also to construe the valence between form and meaning bringing new perspective on the ontology construction.
Construction of a multi-point of view Semantic Web
Keywords : Semantic Web, Ontology, Ontology Matching, Viewpoints.
This work is carried out within the context of Thanh-Le Bach' s PhD.
The objective of this thesis is to allow to construct and use a semantic web in a heterogeneous organization, comprising various sources of knowledge and various categories of users and requiring the management of multiple ontologies or of a single ontology organized in multiple viewpoints.
During this year, we continued to study the state of the art on current alignment techniques. We contributed to the workpackage 2.2 ``Heterogeneity'' of the European Network of Excellence, Knowledge Web ( http://knowledgeweb.semanticweb.org/). A new algorithm for matching/aligning two ontologies is being studied, as extension of the previous one dedicated to RDF(S) . We try to build it based on the ontology structural information and all the other available information possible in ontologies. The new algorithm will focus on the matching/aligning for two ontologies represented in OWL, the web ontology language recommended by W3C. In the first stage, the algorithm will deal with ontologies represented in OWL Lite. So the available information in an OWL (Lite) ontology will be extracted from 40 OWL (Lite) primitives (constructs or elements) such as owl:class, owl:cardinality... and from individuals described in the ontology.
We also proposed a representation of C-VISTA model in OWL. C-VISTA model is a model proposed by Myriam Ribière for representing multiple viewpoints, so it allows to represent multi-viewpoints ontologies . The result is an OWL ontology representing C-Vista model. It will be used to construct a new OWL multi-viewpoints ontology. This OWL ontology will be validated in the future (in the real-world uses) and will be modified if necessary.
As further work, the new algorithm for matching/aligning OWL ontologies will be improved, implemented in Java and tested with ontologies in the framework of Ontology Alignment Contest http://co4.inrialpes.fr/align/Contest/ and other test ontology pairs at I3CON (the Information Interpretation and Integration Conference) http://www.atl.external.lmco.com/projects/ontology/i3con.html. The new algorithm will be also adapted for matching/aligning multi-viewpoints ontologies.
Management of Corporate Semantic Web Evolution
Keywords : Ontologies, Annotations, Life cycle, Evolution, Versioning, Corporate semantic web.
This work is being carried out within the framework of Luong Phuc Hiep's PhD research.
Nowadays, business dynamics and changes in the operating environment often give rise to continuous changes in application requirements, especially in the web-based applications. The web changes at an incredible pace, much faster than a user or even an intelligent web agent can keep up with. Due to the ever increasing complexity, heterogeneity and physical distribution of the information resources, the Semantic Web applications therefore need efficient mechanisms to cope with changes in the environment.
One of our current research approaches is to study the life cycle of a Corporate Semantic Web and the problems related to its evolution: evolution of each component (resources, ontologies and annotations) and evolution of relations among these components.
We aim at studying thoroughly the problems of redundancy and consistency regarding ontology modification, influence on the set of dependent components of an ontology (semantic annotations on these resources, annotated resources with ontology and other dependent ontologies), the problem of version management of some ontologies and annotations and the problem related to cooperative and distributed evolution of ontologies and annotations through the multiple human agents and software. For the first step, we focused on analyzing some related work on ontology evolution. In particular, we have studied the change management for distributed ontologies and some existing methods and tools for ontology evolution as well as the cooperative relations between ontologies and semantic annotations. This work is also a part of the workpackage WP2.3 ``Dynamics'' of the Knowledge Web Network of Excellence.
As further work, we will continue to study ontology evolution and its influence on the resources and annotations depending on the evolving ontology. We also started to work on the metadata evolution and the problem of management of these changed versions of ontologies and annotations.
Semantic Web and E-Learning
Keywords : E-learning, Ontologies, Annotations, Pedagogical Resource Retrieval, Pedagogical Resource Composition.
This work takes place in the framework of Sylvain Dehors's PhD.
Several domains of both topics (Semantic Web and E-Learning) were investigated through an extensive bibliographic research. Below is presented a brief overview of the outcome of this task: it appears that many research fields now deal with e-learning issues, some of them were originally interested in achieving learning through computer use like Intelligent Tutoring Systems and AH. But as new paradigms of learning grow popular, communities dealing with KM are tackling the problem as well, dragged along with others by the universal and irresistible current of the spread of the Web.
This "webization" comes with new ideas such as the Semantic Web, which aims at standardizing and facilitating knowledge sharing. The scientific evidence we've been gathering during this year clearly shows a convergence move towards standardization. Work on standards such as LOM or SCORM appear to be the first step in that direction. One of the questions still raised is whether and how the Semantic Web can play a key role in this process.
As a start to take advantage of semantic web techniques for e-learning, we considered e-learning as a new scenario for knowledge management. By using the semantic search engine Corese, we explored several common e-learning scenarios where learning material was looked upon as a set of learning resources.
In any domain, the first issue raised by using semantic web techniques is to get sufficient processable information to work with (metadata, annotations, resource descriptions, etc.). So our first experiment consisted in building an annotation tool to allow the expression of this information in a machine understandable way, namely RDF, a W3C standard the Semantic Web relies upon.
The expression of that kind of information is tightly related to an ontology, in this case we restricted to model to the pedagogical domain. We used existing models and schemes to build a specific ontology based on LOM and Dublin Core standards. Once combined with the annotation tool, the ontology offers perspectives for model-guided human annotation processes.
And finally we are currently setting up an experiment to evaluate how semantic techniques can improve the common e-learning situation where undergraduate students access course material through the web. This experiment involves a cooperation between two members of ACACIA (Sylvain Dehors and Alain Giboin) and two members of I3S (Catherine Faron-Zucker and Jean-Paul Stromboni) who both teach at ESSI, where the experiment is to take place at the beginning of February 2005.
Semantic Web Services
Keywords : Semantic Web, Web Services.
The ACACIA team focuses on knowledge management solutions based on semantic Web technologies and the last five years resulted in the development of a semantic Web search engine (Corese) enabling us to analyze, query and infer from descriptions in languages of the semantic Web, RDF(S)/OWL, taking into account the ontologies on which they are based. These annotations generally describe documentary resources but, when relying on schemata as the ones advocated in OWL-S, these annotations can describe web services available online (intranet, extranet, open Web) allowing us to automate identification, composition and invocation of these services to provide high-level functionality through dynamic integration.
Thus, we initiated a new activity to specify and implement a generic engine that could discover and compose services allowing applications to plan and execute chains of web services realizing high-level tasks required by a user. The goal is to rely on the semantic search engine Corese to solve queries on descriptions of services, taking into account the ontologies used to characterize them, for instance during the matchmaking of inputs and outputs while forming chains of services.
The problem of planning the composition can be studied from three possible perspectives: assisting manual composition, semi-automatic composition and fully-automatic composition. The problem of running a planned composition will include taking into account potential failures when invoking a service and the discovery/management of possible alternatives.
This work will take place in the framework of the post-doc of Liana Razmerita and, next year, in the framework of the visit of a researcher from Gaston Berger University of Saint-Louis, Senegal, with an AUF grant.
Mobility, Context-Awareness and Privacy
Keywords : Context, Awareness, Privacy, Mobility.
Participant : Fabien Gandon.
Increasingly, application developers are looking for ways to provide users with higher levels of personalization that capture different elements of a user's operating context, such as her location, the task that she is currently engaged in, who her colleagues are, etc. While there are many sources of contextual information, they tend to vary from one user to another and also over time. Different users may rely on different location tracking functionality provided by different cell phone operators; they may use different calendar systems, etc.
We worked on a Semantic e-Wallet aimed at supporting automated identification and access of personal resources, each represented as a Semantic Web Service. A key objective was to provide a Semantic Web environment for open access to a user's contextual resources, thereby reducing the costs associated with the development and maintenance of context-aware applications. A second objective was, through Semantic Web technologies, to empower users to selectively control who has access to their contextual information and under which conditions. This work was initially carried out in the context of myCampus, a context-aware environment aimed at enhancing everyday campus life and empirical results were obtained on Carnegie Mellon's campus .
New scenarios have been envisaged this year through an application to museum tour guides developed for III Taiwan and an application to the assistance of visual deficient submitted in a proposal for a CNRT project.
Support to Cooperative Work: Application to a Health Care Network
This work was performed in the framework of the Ligne de Vie project (see section 7.2). The ACI Ligne de Vie project objective is to develop a knowledge management system for a health care network, so as to ensure care continuity and support to collaborative work of the actors of the network.
This year, our contribution consisted of developing knowledge management tool for a health care network, called Virtual Staff, a tool for processing of patient cases and visualization of collective reasoning using conceptual graphs.
In the hospital, the unity of location and of time allows the doctors to meet as a staff in order to discuss about the decisions to take. In a health care network, the Virtual Staff aims to be a collaborative work supporting tool, allowing the real time update and history of therapeutic decisions. As an electronic board where each one can note information readable by the other members of the team, it constitutes a discussion support that may be synchronous (if the participants take part to the discussion at the same time or in the same place) or asynchronous (if each one accesses it at the moment appropriate to him/her). Starting from the patient's health problems, the members of the team will formulate diagnostic hypotheses and proposals for a treatment. Via this Virtual Staff, the care team will connect the various elements of the patient record useful for the discussion, and thus will converge in an asynchronous way towards the identification of new health problems and of new therapeutic actions.
We suppose that several patient cases will be created asynchronously by multiple users. This requires to visualize and store all information about the patient case, including multi-user and temporal aspects. Another objective of Virtual Staff is to collect solved cases in a special repository and to reuse them for retrieving past patient cases similar to a new patient case to solve. This kind of reasoning - that is based on Corese semantic search engine - is similar to case-based reasoning.
The Virtual Staff relies on a medical ontology Nautilus represented in RDFS and generated in 2003 from the Nautilus medical database by Acacia research team . The Virtual Staff enables to visualize cooperative reasoning through cooperative building of graphs based on the SOAP (Subjective, Objective, Assessment, Plan) model used in medical community and on the QOC (Questions, Option, Criteria) model used by CSCW (Computer-Supported Collaborative Work) community for support to design rationale and decision-making support .
To reach platform-independence, Virtual Staff is completely implemented in Java. The prototype was validated by our industrial partner Nautilus SARL.
Ontology for Manufacturing
This work took place in the framework of the post-doctoral visit of Jiehan Zhou on a semantic manufacturing knowledge management system
We proposed an ontology-driven solution for shared manufacturing knowledge understanding . We analyzed knowledge management requirements from knowledge-intensive manufacturing, designed a semantic manufacturing knowledge management system, developed common manufacturing knowledge ontology and implemented a semantic knowledge query system with the semantic search engine Corese , .