To quote the job announcement:

Current Semantic Web reasoning systems do not scale to the requirements of their hottest applications, such as analyzing data from millions of mobile devices, dealing with terabytes of scientific data, and content management in enterprises with thousands of knowledge workers.

The Large Knowledge Collider (LarKC, for short, pronounced “lark”), a platform for massive distributed incomplete reasoning will remove these scalability barriers. This will be achieved by:

• Enriching the current logic-based Semantic Web reasoning methods with methods from information retrieval, machine learning, information theory, databases, and probabilistic reasoning,
• Employing cognitively inspired approaches and techniques such as spreading activation, focus of attention, reinforcement, habituation, relevance reasoning, and bounded rationality.
• Building a distributed reasoning platform and realizing it both on a high-performance computing cluster and via “computing at home”.

The project consortium is an interdisciplinary team of engineers and researchers in Computing Science, Web Science and Cognitive Science, well qualified to realize this ambitious vision. The Large Knowledge Collider will be an open architecture. Researchers and practitioners from outside the consortium will be encouraged to develop and plug in their own components to drive parts of the system. This will make the Large Knowledge Collider a generic platform, and not just a single reasoning engine.

What more shall I say than: Way to go LarKC!

I tried to figure out some additional information about the project.

What I’ve found is this blog entry. A colleague of mine discovered a presentation given at a workshop.

If you are interested in research related to information retrieval / search and semantic web you might be interested in these two meetings:

• Semantic Search 2008 Workshop
• Exploiting Semantic Annotations in Information Retrieval

Edit: Knud Möller wrote a nice wrap-up about the session in which we both presented our research.

Currently, there is a lot of work going on in the development of query languages for the Semantic Web, see SPARQL for example. With these query languages you specify a query and an interpreter will return you those parts of an RDF graph exactly matching your query. This is a different search paradigm that we are used from search engines like Google. There, you enter a query and you will get results similar to that query, with the documents ranked highest that are most similar to it. Ranking by relevance is something not possible in SPARQL, as every results returned is equally relevant.

My personal QOTD on this interview comes from Ian Davis:

Yeah, I like all that. But I want it bigger and I want anyone to be able to do it

This kinda reminds me of …

the Semantic Web …

as it is now.

And if you are using FOAF check out my profile under: http://scheir.net/foaf-pscheir.rdf (watch the video and know what I mean )

[Via glück auf!]

• Collaborative Protégé: a new extension of the existing Protégé system that supports collaborative ontology editing. In addition to the common ontology editing operations, users can annotate both ontology components and ontology changes. There is support for searching and filtering of user annotations based on a variety of different criteria. Two types of voting mechanisms have been implemented to allow users to vote on change proposals. Collaborative Protégé can be used with the Protégé-OWL editor, the Protégé-Frames editor, and client-server Protégé. Please note that Collaborative Protégé is still considered a prototype that is under active development. Your feedback is most welcome! An installation guide, a user’s guide, an online demo, and more documentation are available on a separate Web page.
• DataMaster: a new plug-in that allows users to import schema structure and data from relational databases into Protégé. DataMaster is documented on the Protégé Wiki.

Check out the release notes.

The biggest change is that Swoogle?s IR index is now updated incrementally, as new or modified Semantic Web documents are processed. When Swoogle processes an RDF document, it analyzes it to extract metadata, and then adds or updates the metadata in Swoogle?s database as well as (re-) indexes information about the document in Swoogle?s IR engine. Previously, these information in the database was updated as documents were found but the IR index was regenerated periodically in an off line batch process. Consequently, the two were not completely synchronized. They are now, at least on a daily basis.

A Categorization Scheme for Semantic Web Search Engines (Esmaili & Abolhassani, 2006)

In addition check at the references on the Search on the Semantic Web page (of course there is some overlap - search on the semantic web is a young discipline ).

• Requirement 1—standard formats: … Using standard formats is preferred, wherever possible, because the investment in marking up resources is considerable and standardization builds in future proofing because new tools, services, etc., which were not envisaged when the original semantic annotation was performed may be developed. Compliance with standards also frees companies from the constraints of proprietary formats when choosing knowledge management software. It is the activity of the W3C in developing and promoting international standards for the SemanticWeb that has convinced us that this route is worth following in knowledge management. Two types of standard are required, standards for describing ontologies such as the Web Ontology Language OWL and standards for annotations such as the W3C’s RDF annotation schema.
  Many of the reviewed tools already use W3C standards.
• Requirement 2—user centered/collaborative design: … Annotation can potentially become a bottleneck if it is done by knowledge workers with many demands on their time. Since few organizations have the capacity to employ professional annotators, it is crucial to provide knowledge workers with easy to use interfaces that simplify the annotation process and place it in the context of their everyday work. A good approach would be a single point of entry interface, so that the environment in which users annotate documents is integrated with the one in which they create, read, share and edit them. System design also needs to facilitate collaboration between users, which is a key facet of knowledge work with experts from different fields contributing to and reusing intelligent documents. …
  … More attention needs to be paid to build in or plug-in semantic annotation facilities in commonly used packages to encourage knowledge workers to view annotation as part of the authoring process not as an afterthought, and also to supporting annotation in collaborative environments, …
• Requirement 3—ontology support (multiple ontologies and evolution): … annotation tools need to be able to support multiple ontologies. For example, in a medical context, there may be one ontology for general metadata about a patient and other technical ontologies that deal with diagnosis and treatment. … In addition, systems will have to cope with changes made to ontologies over time, such as incorporating new classes or modifying existing ones. In this case, the problem is ensuring consistency between ontologies and annotations with respect to ontology changes. …
  … Ontology maintenance, which directly affects the maintenance of annotations, is poorly supported, or not supported at all, by the current generation of tools. This perhaps reflects the intended user groups; with the assumption being that knowledge workers will use existing ontologies rather than editing or creating them. … A genuinely integrated semantic annotation environment should give the user automatic support for ontology maintenance, for example, using text mining methods to suggest new classes as they emerge in documents and spotting inconsistencies between new and existing annotations. …
• Requirement 4—support of heterogeneous document formats: … Documents will be in many different formats including word processor files, spreadsheets, graphics files and complex mixtures of different formats. This presents a technical challenge rather than a research challenge, but dealing with multiple document formats is a prerequisite for integrating annotation into existing work practices.
• Requirement 5—document evolution (document and annotation consistency): Ontologies change sometimes but some documents change many times. … What should happen to the annotations on a document when it is revised, poses both technical and application specific questions. … Annotation environments need to help knowledge workers maintain appropriate annotations as documents change.
  The survey did not discover any concerted work on these lines.
• Requirement 6—annotation storage options: The Semantic Web model assumes that annotations will be stored separately from the original document, whereas the “word processor” model assumes that comments are stored as an integral part of the document, which can be viewed or not as the reader prefers. …
  … However, separate storage of annotations has advantages for KM. … It also makes it easy to produce different views of a document for users with different roles in an organization or different access rights, thus facilitating knowledge sharing and collaboration. We therefore argue that separate storage is the better model, even when extra overheads are required to maintain links between a document and its annotations.
• Requirement 7—automation: Another aspect of easing the knowledge acquisition bottleneck is the provision of facilities for automatic mark-up of document collections to facilitate the economical annotation of large document collections. To achieve this, the integration of knowledge extraction technologies into the annotation environment is vital. …
  Language technologies present usability challenges when deployed for knowledgeworkers since most are research tools or designed for use by specialists. … In addition to the usability challenges there are also research challenges, among which we have highlighted the extraction of relations as important for semantic annotation.

• Decoupling the process of engineering from that of exploiting the Semantic Web. NGSWA assume that they operate in an environment characterized by large scale, distributed semantic markup.
• Operating with heterogeneous semantic markup and multiple ontologies. NGSWA have to deal with heterogeneous semantic markup.
• Openness with respect to semantic resources. NGSWA allow to add new sources or integrate new ontologies
• Scale more important than quality. While a lot of the emphasis in first-generation tools was on quality, NGSWA move away from traditional quality centered expert systems, just as the Web differentiated itself from hypertext, by allowing broken links.
• WWW – We Want Web! Early Semantic Web applications are far more similar to the classic knowledge-based systems, than to the Semantic Web applications of the future. NGSWA try to bring the Semantic Web closer to the Web and also integrate Web Services in their functionalities.
• From intelligent applications to harvesting collective intelligence. In NGSWA intelligence is also a byproduct of operating with large amounts of data. The users act as catalysts in deriving value from collectively gathered, tagged and shared semantic data, thus using the system to harvest collective intelligence.

See also my previous post on Characterizing Semantic Web Applications.

Beside the detailed survey on Semantic Annotation they also present seven requirements for Semantic Annotation in a document centric approach to Knowledge Management.

• Language Technologies and the Evolution of the Semantic Web (Motta and Sabou 2006a)
• Next Generation Semantic Web Applications (Motta and Sabou 2006b)

These are the (my) key-points of the second paper (Motta and Sabou 2006b).
Properties of Next generation Semantic Web applications (NGSWA) are:

• Semantic data generation vs reuse. NGSWA are designed to operate with the semantic data that already exist. In other words, they worry less about bootstrapping a Semantic Web, than about providing mechanisms to exploit available semantic markup.

• Single-ontology vs multi-ontology systems. NGSWA can consume any number of ontologies at the same time. It does not make much sense to make a ‘closed domain’ assumption.

• Openness with respect to semantic resources. NGSWA take into account RDF data available from a particular Web site, in response to a request from a user who wish to use them.

• Scale as important as data quality. NGSWA are designed to operate at scale.NGSWA do not require any extra effort to bring in new sources.

• Openness with respect to Web (non-semantic) resources. NGSWA integrating data acquisition mechanisms in their architecture.

• Compliance with the Web 2.0 paradigm. NGSWA need to provide mechanisms for users to add and annotate data. NGSWA support user annotation are still rather primitive, and better tools are badly needed.

• Open to services. NGSWA seamlessly integrate scraping services into their data acquisition architectures.

This is the private research blog of Peter Scheir, dealing with topics related to Ontologies, the Semantic Web and Web 2.0 in the context of information retrieval and technology enhanced learning. While I am lucky enough to deal with some of the topics presented here during my work at the Know-Center, this blog is not associated with the Center or my work there.