This poster discusses the problem of making the contents of heterogeneous databases semantically interoperable on the web. We argue that semantic interoperability is needed in order to combine museum collection data logically and to provide the museum visitor with useful search and navigation services into the semantically rich cultural contents. A demonstrational implementation that combines two Finnish museum databases is discussed.
The distribution of collection objects in museums at different locations creates an obstacle to information retrieval, for both the public and for researchers. It should be possible to use the collections as if they were in a single database. In this poster, based on the paper , we address the problem of making existing distributed collection databases mutually interoperable on the semantic level. We argue that semantic web technologies offer a promising approach to facilitating homogeneous, semantic information retrieval based on heterogenous databases on the web.
In order to realize this goal, a research and development project* is being carried out at the Helsinki University Computer Science Department and the Helsinki Institute for Information Technology (HIIT). As a case study, the collection databases of the Espoo City Museum and the National Museum of Finland are used. These databases use different DBMSs (SQL Server and Ingres), have different relational schemas, and are located physically in different cities. The systems are representatives of two major camps of Finnish museum information systems: the Antikvaria group (the Espoo City Museum) and Musketti (the National Museum). For the sake of simplicity, the implementation of the system is first restricted to only one part of exhibit collections, the textiles.
The traditional approach in publishing collections on the web is to use keyword-based search for the collection database. The user types in keywords and the system generates hit lists matching with the given words. This approach is used, e.g., in the Australian Museums Online system**. In our view, such functionality is useful especially to an expert user looking for some specific objects. The goal of the ordinary museum visitor is, however, usually something quite different from trying to find certain objects. One would rather want to learn about the past and experience it with the help of the collections. In physical exhibitions the cognitive museum experience is often based on the interesting semantic and thematic combination of exhibits and their contextual information. The same principle applies to WWW exhibitions, as well.
To take a step towards this ambitious goal, the Finnish Museums on the Semantic Web (FMS) system transforms collection databases into a virtual semantic web space. Its pages are linked with each other with semantic links that are useful for finding information based on its content. The idea is to offer to the user a semantic browsing and searching facility in the combined collection knowledge base . This facility is implemented by a piece of server-side software, called Ontogator. When the user views the exhibition entry page with a web browser, Ontogator dynamically generates WWW pages with links to other pages of interest. The FMS home page is the single entry point through which the user enters the virtual museum collections' WWW space.
|Figure 1: The architecture of the Finnish Museums on the Semantic Web system.|
The general architecture of the FMS system is presented in figure 1. Museums join the system by first generating exhibit cards in RDF from the database. They conform to an annotation ontology. This data must be checked by a human editor before publication and is annotated according to the exhibit ontology used. In our first experiment, Protégé-2000*** has been used for the task. The metadata is placed in a public directory on the museum's WWW server. In this way, the museum can easily and completely control of the information it wants to publish. The museum does not need to allow the FMS system access their internal database system through the web, like in traditional tightly coupled approaches****. The collection data of different museums is read in by the web crawler of the FMS system and combined into a global RDF database. Based on this graph, Ontogator dynamically generates the collection WWW space for the user's web browser.
Ontogator will provide the user with the following semantics-based facilities.
View-based filtering Ontogator shows the multiple ontologies used in annotating collection data, such as ObjectType, Material, etc. By selecting ontological classes from these hierarchies, the user can express the search profile easily in the right terminology. For example, by selecting ObjectType=carpet and Material=silk, silk carpets are found. This view-based idea to information filtering is adapted from the HiBrowse system developed for a bibliographical information retrieval system . Using the system is based on the metaphor of opening directory folders - the idea used in Windows Explorer and in many other systems.
Topic-based navigation Ontogator supports topic-based navigation according to the underlying idea of Topic Maps . The creation of semantic links between topics of interest is based on 1) the collection domain ontologies (classes and their relations) and 2) on actual collection data (instance data). The links give the user contextual and pragmatic information about the objects in the collection.
Ontological search engine for Finnish A search engine is being developed for generating hit lists in the same fashion as search engines on the WWW. However, our engine will understand and make use of the semantic relationships between keywords. The conjugation of Finnish words is also taken into account of.
|Figure 2: A semantic browser for the image database of the Helsinki University Museum. Three ontological views to the material are opened on the left (Place/Paikka, Person/Henkilö, Event/Tapahtuma). On the right, an image is selected for viewing and below it semantically related other images are recommended by the system.|
The actual implementation of the system is underway. To illustrate the use of the FMS system, figure 2 shows the user interface of our first experimental implementation of Ontogator. This system  was originally created for an image database of the Helsinki University Museum, and will be developed further in the FMS. The system implements view-based filtering and facilitates topic-based navigation in a restricted sense by recommending semantically related images. On the left in the window, the user may choose different ontological views of the images related to the promotional ceremonies of the University of Helsinki. Here the ontologies of ''People'', ''Places'', and ''Events'' are open. The ontologies show the user the concepts that will be needed when searching for images. This is necessary, because the end-user might not be familiar with the promotion-related concepts. By opening the ontology hierarchies, the user chooses the ones that are of interest, and the browser shows all the images that fit all the chosen concepts. On the right, an image found by this method is shown. Underneath it, the system displays a number of related recommended images that are connected semantically with the main image in some way. The system finds such images automatically based on the underlying RDF database. For example, in the recommended images, the same persons may appear as in the main image, but in another context. By browsing the collections according to the associations, the user may make a journey into the world of promotional events and images.
Our work is funded by the National Technology Agency Tekes, Nokia, TietoEnator, the Espoo City Museum, and the Foundation of the Helsinki University Museum, and the National Board of Antiquities.
**** E.g., the Z39.50 protocol, http://lcweb.loc.gov/agency/