Project Title: Cultural Heritage in an Interactive Multimedia Environment Project Acronym: CHIME Principal Investigator: Lynda Hardman Classification a) Opleiding en Cultuur, b) Gebruiker en kennisveredeling Composition of the Research Team [[make table]] CWI generation, 1 post doc (Frank Nack?), 1 OIO TUE User characteristics, learning, 2 oio's?? SWI MM ontologies, 1 OIO Rijksmuseum data, Koninklijke Bibliotheek data National bodies such as the Rijksmuseum and Koninklijke Bibliotheek maintain repositories of cultural heritage that are mostly inaccessible to the general public and experts distributed around the world. Large investments are being made to capture as much of this information as possible digitally, and projects have been carried out to give broader access to the digitised material. There are two limitations with the current approaches, however. First, that they focus on a single user, e.g. novice user in the Rijksmuseum ARIA system and expert user in the Rijksmuseum AdLib database, and that the system is uni-directional, i.e. "experts" input information into the system and "users" query this information. The goal of CHIME is to remove these restrictions and allow information to be presented to a range of users in a suitable way and to allow all users to add their own information to the repository, while respecting the integrity of the original historical sources. This allows a decentralised approach to the enrichment of the information in the repository by all its users to the benefit of all its users. An example scenario to illustrate the expected usage of such a system is the following. A class of students are involved in a project (e.g. "studiehuis") and have been given an assignment on the topic of Dutch history. In addition to querying the CHIME system to find pertinent information, they are also able to add their own, private, remarks to the repository. Relevant multimedia data, such as figures, video material, photos, newspaper articles, spoken commentary can also be added to the repository. When the assignments are completed they are reviewed by the teacher, who is able to include reviews of the assignments in the repository. In the final stages of the assignment, each group gives a presentation of their work, which is recorded on video. The teacher then includes the videos of the best groups in the repository. To realise this scenario, the system needs to take into account different access levels, privacy issues and the varying degrees of trustworthiness of the data entered by different users. In addition, research has to be carried out in three different areas. First, the creation of ontologies to describe and relate the key concepts of each of the components in the system to enable a shared understanding of these concepts throughout the system. Second, the relationship between ontologies and multimedia. Third, the evolutionary development of the ontologies and the information repository. Specific research questions to be tackled are: At which granularity level do users need to be modelled? For example, is a user group level sufficient, or do individual user profiles need to be created? How do we reduce the required amount of user-effort needed to create these profiles? Can common ontology-based models and tools support the development, interchange and maintenance of user profiles? Which characteristics of a user need to be modelled in order to facilitate retrieval and presentation of information, and also to assist the user in entering information in the repository. For example, which parts of the annotation ontologies need to be made available to which user? In addition, the user characterization will play a key role in deducing the trustworthiness of the entered annotations. For example, information entered by a museum curator is more trustworthy than that entered by a student. Which levels of trustworthiness of data and metadata (note that the distinction between these is relative) are required? How to extend the current state of the art on the Semantic Web in a way that makes user, domain and presentation-specific ontologies applicable for describing multimedia content? Which other types of metadata need to be assigned to the information in the database to allow user-centered filtering and presentation generation to be automated? Do media items need explicit annotations that add information relevant to the discourse model, such as "this is an explanatory text", "this video fragment is an example of that concept in the subject domain"? How to relate ontologies (domain ontology and multimedia ontology) to make sure that terms from multiple ontologies can be used within a single instance? How to design repositories that allow a decentralized, evolutionary growth of the media content, the metadata and the associated ontologies? What tool support is needed to support such a collaborative approach? How to integrate privacy concerns, access policies, trust levels, versioning control etc. into these tools? How to assist manual annotation using automatic feature detection techniques? The results of the project will lead both to scientific publications from the research results and prototype implementations, building on the presentation generation framework created at CWI during the first phase of ToKeN2000. In addition we expect the results of the project to influence the construction of the Semantic Web. In particular, at the time of writing, a large amount of international effort (European and American) is being devoted to the construction of the Semantic Web. To date this effort has been media independent, in the sense that it is predominantly text-centric. To exploit the Web to its full potential, multimedia needs to become an integral part of the Semantic Web infrastructure. While the primary research goals of the project are not targetted at the development of the Semantic Web, the results from the project will be relevant to it. The research methods used will involve an interdisciplanary approach, combining knowledge acquisition techniques from AI and user driven design from HCI. [[develop software prototypes and ontologies, talk to users, literature studies.]] The research fits into the ToKeN2000 framework in a number of ways. First, it concentrates on media-centric problems already relevant to institutions such as the Rijksmuseum and Koninklijke Bibliotheek, not to mention future problems with national archiving of electronic media. Second, it enriches the information search process not by presenting results as an ordered list, but by combining relevant information in a coherent hypermedia presentation, allowing details to be requested through links and otherwise presenting the most appropriate material in the most appropriate form. Third, the nature of the research requires an interdisciplinary team. Fourth, the results can only be satisfactorily demonstrated through the implementation of a working software environment to test out research hypotheses. Fifth, [[ need to position proposal w.r.t. other work - RTIPA, Dynamo, I2RP proposal]] Research results can be made available to a broader community in a number of ways. In particular, CWI has strong links with W3C's working groups and hopes to continue in its tradition of influencing relevant developments with its research results. A Semantic Web initiative is currently being set up and results from the project will be relevant to this initiative. Expertise from the knowledge representation community, in particular from the SWI group, can also feed back into developments in RDF and RDFS at W3C. Software components from the project are likely to be usable by a broader community, in particular the components relating to the Semantic Web. The research has potential application in broad areas. Not just in the museum domain, although this is a domain particularly suited to [[added value of building up repository for others to use and to add to...attaching semantic information .]] Any content provider interested in providing their content to different users, through different networks on different hardware platforms is a potential user of the research results. For example, news providers, such as NOS and ANP, want to be able to deliver their content rapidly to many different end-users. Role of partners is CWI adaptive hypermedia presentations, multimedia annotations and semantics. TUE user profiling, learning environments, hypertext access to semi-structured databases. SWI knowledge elicitation, ontology tooling.