Integrating external sources in a corporate semantic web managed by a multi-agent system

	Download document (): Word, PDF
	Download article presentation (): PPT+ZIP
	Download wrap up session presentation (): PPT+ZIP
	Download "fake panel" session, presentation of opposit opinion (): PPT+ZIP

“D.A.I. & S.M. for KM”a synergy of complementary domains and challenges

• Other K.R. languages existed before but none of them made it to the real world ; some part also matured (ontology)
• SW is a real-world application (the Web) for K.R.
• the SW is also about standardization and diffusion effort for semantic representation on the Web.

• High level programming and design paradigm that reduces conceptual gap between description of our reality (problems and envisioned solutions) and the description used in the modeling and implementation framework.

• Distributed A.I. offers a paradigm and architectures to deploy and map over distributed knowledge spaces
• Virtual organizations can reflect, and integrate with human organizations

• The use of abstract categories shows up in a lot of work
• XML is not enough, the machine does not understand <car /> any more than “car” ; need for ontologies and SW
• True both for the open Web and for the intrawebs
• Even if the human brain representation is completely different of the ones (D)AI is using, if our symbolic systems can simulate the inferences we want using ontologies then why not use them?

• Contrary to previous attempts, the “ontology” object and its problematics are recognized and being addressed.
• There is an effort in trying to build standard top ontologies (SUMO), and domain ontologies that can be reused and extended by organizations.
• No imposed standards, make them available and show benefits to everyone  ; otherwise it will not happen

• Semantic content or statistic content?
• A lot of low-quality ontologies on the Web but they will disappear with time / hope they won’t harm the domain
• The linguistic / semiotic level is too often mixed-up with the conceptual structures and representation themselves ; need for separation and development of this level.
• Problem of pragmatic use of terms / signs and interpretation not really addressed
• Content and semantic are largely underestimated, tools and methods are too much emphasized
• You have to go through a period of chaos before you reach a stable situation
• Transition period: going to double web before going where everything is in the markup.
• SW initiatives also provide rules, constraints on how it should be done i.e. it is more than a simple syntactic sugar

• “Build small but viral” Tim Burners-Lee
• 80/20 rule for dissemination. Let the demand for the rest come after
• Choose the right domain to build and demonstrate ontologies (e.g., services, processes, interaction protocols)
• Then tend toward a maximum of expressiveness and overlap with other existing ontologies
• Top ontologies and standard domain ontologies are vital to foster this convergence and make the compatibility possible.

• Layers? The semantic web cake.
• Components? But too much anarchy would be dangerous.
• Top ontology (essential) + hierarchies of extensions and management of overlaps between extension
• Semi-automatic mapping for relevant parts

• Exact answer to my query? Improving search mechanisms?
• Mechanisms to reduce number of answer?And what if there really are  1,000,000,000,000,000 answers
• Real improvements? Not ambiguity.
• Only as good as the expressivity of the ontology.
• Need more weighting / fuzzy ? No, just sub-type of Ont. K.
• Pornography ?
• Hmm let say… “Multimedia”

• Use formal knowledge to evaluate some quality (e.g., coherence) and security (e.g. access policies)
• Use for filtering and ranking
• Some solution of K.M. (e.g., peer review, trust authorities and (acquaintance) networks)

• Large organizations with intranet = interesting special case
• Intraweb application are a good domain of application (information systems and workflow)
• Problem of burden, separation of concerns in the company (worker vs. K Manager)

• Virtual enterprises
• Company merging

• Corporate internal ontologies
• Top ontology ex: SUMO then extension with domain ontologies

• “Too complex to be shown to a user”
• No reason to show it to a user

• Forms are not usable for every interactions
• More intelligent interfaces using semiotic levels
• “We focus, and interfaces should focus with us”
• Pragmatic aspects of language in interfaces

• Some of it manually (e.g. building an ontology)
• Some of it from (semi-)automatic process
• Pragmatic aspect of the interpretation of the content of the Web

These schemata are an extension of the schema found in the conclusion of my PhD dissertation:

It underlined that the fields chosen for my research are linked by a very strong complementarity:

• Utility of ontologies for the agents: the communication between agents is based on speech acts provided by an ontology, their inferences exploit semantic spaces defined by ontologies to support distributed mechanisms, etc.
• Utility of ontologies for the semantic Web: the realisation of the schemas defining the vocabulary of the annotations calls upon methods of ontology design, the new inferences of research in the bases of annotations exploit the properties of theontology, etc.
• Utility of the agents for ontologies: multi-agent architectures can be used for the realisation of collective software, for setting up and maintaining the ontological consensus, the agents can manage the gateways between various domains and their respective ontologies, etc.
• Utility of the agents for the semantic Web: the agents propose a suitable paradigm for search systems handling distributed annotation bases, the multi-agent architecture can also be used for the deployment of semantic Web services, etc.
• Utility of the semantic Web for the agents: the W3C proposes standardised languages for the Semantic Web which can be used as agent communication languages such as XML and RDF(S), the semantic web allows us to build annotated worlds where information agents can quickly become intelligent actors, etc.
• Utility of the semantic Web for ontologies: the W3C proposes standardised languages for the Semantic Web which can be used for the exchange and the extension of ontologies, the Semantic Web is also a perfect ground of application for ontologies, etc.

This shows that the application of distributed artificial intelligence to the knowledge management problems is a real and challenging research field with a promising future that calls for a dedicated research and development community.