An Ontology Server for Agentcities.NET

Transcription

1 AGENTCITIES TECHNICAL NOTE Agentcities Task Force Technical Note actf-note-00008, 5 Noveber, 200 Authors: Monica Duke, UKOLN, University of Bath Manjula Patel, UKOLN, University of Bath Copyright 2003 is retained by the Authors and/or their respective organizations. The content is the sole responsibility of the authors and ACTF takes no responsible for its correctness or fitness for purpose. The authors are also responsible for ensuring that this publication does not violate copyright agreeents applying to the content in whole or in part. Coents and requests should be addressed to tech-editor@agentcities.org. This docuent and the inforation contained herein is provided on an "AS IS" basis and THE AGENTCITIES TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Status Final This version: Latest version: Abstract Within this six onth deployent project[] we have concentrated on taking forward the ideas and systes developed in a nuber of initiatives in which UKOLN has been involved, chiefly aong these the EU-funded DESIRE[6] and SCHEMAS projects[7], the UK MEG Registry project[5] and the Dublin Core Metadata Initiative[5]. All of these projects explored approaches to declaring and sharing etadata vocabularies using RDF Scheas[8]. We have adapted software for a etadata vocabulary registry to serve as an ontology server which can be queried by agents on the Agentcities.NET network. The contents of the server coprises etadata vocabularies which ay be regarded as siple fors of ontology The Authors

2 Contents Agentcities Technical Note... Introduction Ontologies and Metadata Vocabularies Ontology Description Languages RDF Schea DAML+OIL DAML+OIL RDFS(FA) Ontology Servers and Metadata Registries The SCHEMAS Metadata Registry BT s Ontology Server The Dublin Core Metadata Initiative s Registry Other Initiatives The MEG Registry The MEG Registry odel of etadata vocabularies The UKOLN Ontology Server Web Interface The UKOLN Agent Platfor Overview of functionality The Server Agent Server Ontology Interrogating the Server Agent The GUI Agent The Coand Line agent Behaviours Conclusions Acknowledgeents References... 8 Appendix: The MEG Registry Data Model The Authors 2

3 Introduction This is a report on the work carried out between st Septeber 2002 and 28 th February 2003 at UKOLN, as part of the European Coission funded 5 th Fraework IST project Agentcities.NET [4]. UKOLN was awarded a grant under the Deployent support progra, a "series of grants to support independent new innovative exploratory work related to the Agentcities.NET network. The intention is to enable ebers to connect their existing or new agent systes to the Agentcities network and carry out exploratory ini-projects - leading to innovative ideas, technology developent and new larger scale collaborative projects." UKOLN [3] is a centre of expertise in digital inforation anageent, providing advice and services to the library, inforation, education and cultural heritage counities. UKOLN is involved in any standardization activities, including the Dublin Core Metadata Initiative (DCMI)[5]; the Research and Developent tea at UKOLN has taken part in several EU projects including DESIRE[6] and SCHEMAS[7]. The ai of this project is to investigate the support of autoated querying of etadata vocabularies by agents, to acquire the seantics associated with specific etadata ters. The approach taken is that of using a registry within which etadata vocabularies are expressed and through which they are counicated. In a registry environent, individual ters as well as whole vocabularies can be investigated by agents. The registry supports the discovery, sharing and re-use of vocabularies, facilitating the convergence of vocabularies (or ontologies), in particular for specific doains. The hope is that alignent in this way will iprove the prospects of interoperability of systes in specific sectors. 2 Ontologies and Metadata Vocabularies Ontologies provide a coon vocabulary of an area and define, with different levels of forality, the eaning of the ters and the relations between the. They ai to capture doain knowledge in a generic way and provide a coonly agreed understanding of a doain, which ay be reused and shared across applications and groups [0]. Ontologies are used by people, databases, and applications that need to share doain inforation. There are several other definitions and typologies of ontologies; for an overview [0, ] are good sources. Soe definitions ay follow fro the way that ontologies are built and used; distinctions are ade between lightweight and heavyweight ontologies, where taxonoies are considered to be one of the forer, whereas the latter kind of ontologies would be expected to include axios. For exaple Sowa [2] defines a terinological ontology as "an ontology whose categories need not not be fully specified by axios and definition". WordNet [27] is an exple of such an ontology. Other distinctions are based on the kind of languages used to ipleent ontologies, such that soe ontologies are rigourously foral if they are defined in a language with foral seantics, theories and proofs (e.g. of soundness and copleteness). Others are only highly inforal being expresssed only in natural language. Soe ontologies are intended to be reusable across doains but several are specific to a doain. Knowledge in ontologies is ainly foralized using five kinds of coponents: classes, relations, functions, axios and instances. For a description of these coponents refer to [0]. However, in this project we are concerned with only a specific type of siple ontology, referred to in the SCHEMAS project as a vocabulary[3]: "In our usage, the ter evokes a seantically rich dictionary environent, with pointers to related ters ore than just a flat word list. (Another coon synony for vocabulary is eleent set. Siilarly, though we prefer to speak of etadata "ters", the ter "eleents" is a close synony.) " Further, the SCHEMAS project developed the notion of an Application Profile[9] which is a type of etadata vocabulary that draws on canonical vocabularies and custoizes the for local use. The precise use of the ters vocabulary and application profile and how they are odeled in our work will be expanded on in section The Authors 3

4 Ontology Description Languages Seanticweb.org [25] provides an encapsulation of the history of the representation of ontologies on the Web. More recently the OWL Web Ontology Language[22] is being designed by the W3C Web Ontology Working Group[9] in order to provide a language that can be used for applications that need to understand the content of inforation instead of just understanding the huan-readable presentation of content. OWL facilitates greater achine readability of web content than XML, RDF and RDF Schea[8] by providing an additional vocabulary for ter descriptions. The OWL language is a revision of the DAML+OIL web ontology language incorporating learnings fro the design and application use of DAML+OIL[36]. 2.. RDF Schea The Resource Description Fraework (RDF) is a general-purpose language for representing inforation on the Web. The RDF Schea specification [8] describes how to use RDF in order to describe RDF vocabularies DAML+OIL DAML+OIL [2] is a seantic arkup language for Web resources. It builds on earlier W3C standards such as RDF and RDF Schea, and extends these languages with richer odelling priitives. DAML+OIL provides odelling priitives coonly found in fraebased languages. A DAML+OIL knowledge base is a collection of RDF triples. DAML+OIL prescribes a specific eaning for triples that use the DAML+OIL vocabulary 2..3 DAML+OIL The Web Ontology Language OWL [22] is a seantic arkup language for publishing and sharing ontologies on the World Wide Web. OWL is developed as a vocabulary extension of RDFS and is derived fro the DAML+OIL Web Ontology Language[2]. OWL is a language for defining and instantiating Web ontologies. Different subsets of the OWL language are defined, to suit different uses. OWL has been designed for axial copatibility with RDF and RDF Schea, and an OWL ontology is represented as a set of RDF triples RDFS(FA) RDFS(FA)[28] as a sub-language of RDFS introduces a Fixed layered etaodeling Architecture to RDFS, based on a relatively standard odel-theoretic seantics. Therefore, first order languages, like DAML+OIL and OWL, can be built on top of both the syntax and seantics of RDFS(FA). On the other hand, all RDFS(FA) stateents are still valid RDFS stateents, since RDFS(FA) iposes the restriction of stratification on the syntax of RDFS. It is intended to address the 'dual-roles' proble in RDF. RDFS(FA) is designed to be a clean schea layer language (as a sub-set of RDFS), such that it is easy to understand and to use first order logics (e.g. DAML+OIL and OWL/DL) can be built on top of both its syntax and seantics RDFS(FA) is a Seantic Web schea language introducing a UML-like etaodeling architecture to RDFS. Built-in odelling priitives of RDFS are stratified into different strata (or layers) of RDFS(FA), so that certain odelling priitives belong to certain stratus (layers). The seantics of odelling priitives depend on the stratu they belong to. All these strata for the etaodeling architecture of RDFS(FA). Theoretically there can be infinite nuber of layers in the etaodeling architecture, while in practice, four layers are usually described: Stratu 0 (Instance Layer) Stratu (Ontology Layer) Stratu 2 (Language Layer) 2003 The Authors 4

5 Ontology Servers and Metadata Registries As used in the SCHEMAS Project, the ter "registry" refers to a database that harvests various types of etadata vocabularies fro their aintainers over the Web. In response to queries, such a registry should provide ter-level docuentation of definitions and usage along with contextual annotations. It should in effect function as an indexing engine for dynaically updating, erging, and serving up a large corpus of definitions for etadata ters. The context for such a registry is the notion of a Seantic Web where anybody or any organisation can declare a etadata vocabulary and assert a relationship between that vocabulary and any other vocabulary on the Web. 3. The SCHEMAS Metadata Registry The SCHEMAS project developed a etadata registry which was ipleented using the EOR toolkit (Extensible Open RDF toolkit)[37]. An RDF approach offered the potential of a scaleable syste based on a coon data odel (RDF) both for the schea and for the database. The project was looking towards ipleentation of a repository which would be populated with scheas harvested directly fro their aintainers in an open Web environent. However, at that tie software tools for such a solution proved iature and required a level of developent effort beyond that available to the project. In addition the chosen standard for schea specification (RDF Schea) was itself still under developent, and conventions for expressing etadata scheas, in particular Application Profiles[9], were still to eerge. The priary otivation for the work on the SCHEMAS Registry "has been to help huans find out about etadata ters in use -- their official definitions, local variations and extensions, and the various scheas in which they are ebedded. The purpose is to help designers of inforation services discover etadata ters that have already been created or standardized by others and align their own scheas with those of related inforation providers." [8]. However, the longer-ter goal was "to build a corpus of achineunderstandable scheas that can be accessed and processed directly by various software applications" [8]. 3.2 BT s Ontology Server The BT Ontology Server [3] is part of the Agentcities.RTD initiative. The Agentcities Ontology Service is an agent and web application for anaging and accessing DAML+OIL ontologies and can be accessed by agents using open standards (the Agentcities interoperability stack). This allows ontologies to be created, anaged and shared by agents [32]. 3.3 The Dublin Core Metadata Initiative s Registry The Dublin Core Metadata Initiative is an open foru engaged in the developent of interoperable online etadata standards that support a broad range of purposes and business odels. The overall goal of the DCMI Registry Working Group[35] is the developent of a etadata registry providing authoritative inforation regarding the DCMI vocabulary and the relationship between ters in that vocabulary. The group ais to provide an operational registry with both user and achine interfaces over a phased developent period, with the ai of supporting acceptance and use of the DCMI vocabulary and providing an authoritative source of inforation [35]. Work in this initiative is ongoing. 3.4 Other Initiatives Other intiatives within the areas of ontologies, ontology representation, storage and exchange have undertaken reviews of repositories of ontologies: The OntoWeb Technical RoadMap [0] reported on repositories of ontologies, listing soe of the 'best-known repositories'. The ontology repositories that are described include those in which ontologies are ipleented in DAML, Ontolingua and SHOE The Authors 5

6 More recently, the SWAD Europe Project reviewed RDF storage systes [20] including ones that ay include schea and ontological data such as RDF Schea and DAML+OIL. The DAML Repository [30] is a web-accessible catalogue of ontologies expressed in DAML. 3.5 The MEG Registry The Metadata for Education Group (MEG)[4] was fored following a eeting of key UK stakeholders and serves as an open foru for debating the description and provision of educational resources at all educational levels across the United Kingdo. This group seeks to reach consensus on appropriate eans by which to describe discrete learning objects in a anner suitable for ipleentation in a range of educational arenas. Preceding work undertaken in the DESIRE[6] and SCHEMAS[7] projects provided the basis for the MEG Registry Project[5], which adopted a slightly odified data odel as described in the Appendix. The ai of the MEG registry is to provide ipleenters of educational systes with a eans to share inforation about their etadata scheas and to re-use existing scheas. The benefit being a saving of tie and effort currently spent in researching existing scheas and in re-inventing scheas. In the next few sections we describe in soe depth the odels and definitions eployed in the MEG Registry project as they have provided the fraework for our work The MEG Registry odel of etadata vocabularies The registry is based on the following odel of etadata vocabularies or eleent sets: Eleent Sets are owned and aintained by Agencies. Eleent Sets are ade up of Eleents. An Eleent Usage ay: introduce constraints on the value of an Eleent by associating it with one or ore Encoding Schees; introduce constraints on the obligation to use an Eleent (e.g. ake its use andatory) or the occurrence of an Eleent (e.g. whether it is repeatable); refine the seantic definition of an Eleent to ake it narrower or ore specific to the application doain. Encoding Schees constrain the value space of Eleents. An Application Profile defines a set of Eleent Usages of Eleents drawn fro one or ore Eleent Sets. The registry holds inforation on each of the entities and their relationships: Eleent Sets (i.e. on the Eleent Sets as units, rather than on their constituent Eleents), including inforation on their intended scope/area of use and their relationship to other Eleent Sets; the Eleents which ake up those Eleent Sets, including inforation on the seantics of the Eleents and their recoended usage, and any seantic relationships to other Eleents in this or other vocabularies (e.g. the relationship described by the DCMI concept of "eleent refineent" or by RDF Schea as a "sub-property" relation) Application Profiles, including inforation on their intended scope/area of use and their relationship to other Eleent Sets and Application Profiles; the Usages of Eleents which ake up those Application Profiles, including the Eleent used, any prescription of Encoding Schees, and other constraints on eleent use; Encoding Schees, which constrain the value space of Eleents, including inforation on their intended scope/area of use; where an Encoding Schee takes the for of an enuerated list, the values prescribed by that Encoding Schee ay be recorded; the Agencies who own/create/aintain Eleent Sets, Application Profiles, and Encoding Schees 2003 The Authors 6

7 Diagraatically, the relationship between the entities that are represented in the registry is odelled as follows (a ore foral description is available in the Appendix). Eleent Set Agency Eleen t Encoding Schee Value Eleent Usage App Profile The Meg Registry is ipleented as a server based on the RDF toolkit, Redland [6]. The inforation about the above entities and their relationship is stored and ade available in achine-processible forat as RDF scheas. The existing registry API is developed in Perl and supports functions such as querying of the registry through an HTTP interface. The project also provided a tool that could support the creation and subission of etadata scheas in a distributed way, in particular prooting the re-use of eleents and encoding schees as described in [7]. The registry can be queried either through the schea creation tool so as to identify eleents and encoding schees for re-use, or directly through the HTTP APIs. One of the interfaces was intended for browsing and searching through a web browser, and returns HTML encoded representations of the structures and relationships of the eleent sets and related entites, which support easy navigation through the registry. Thus each of the entites (agency, eleent set, eleent, application profile, eleent usage and encoding schea) can be either searched or browsed and the relationships can be explored. A second interface supports queries to search against eleent sets and encoding schees, and returns RDF-encoded data. 4 The UKOLN Ontology Server Recently, we have extended the work done in the MEG Registry project to re-deploy the interfaces to the registry within an agent environent, naely the Agentcities.NET[]. The existing registry software stores inforation pertaining to etadata vocabularies and provides an interface for interacting with the inforation. We have thus transitioned fro a huancentric to an agent-centric environent. We have deployed the MEG Registry software within an agent-enabled environent, ediating counication to the registry of scheas through an agent. The scheas (or eleent sets) are odelled within the Server as outlined in previous sections and in the Appendix. Exploration of the eleent sets is organised around the categories described by the odel, (i.e. agency, eleent, eleent set, application profile, encoding schee and eleent usage) The Authors 7

8 Web Interface Independent of the agent interface, the Server can also be explored through a web interface, which is linked fro the web page: The following screen shots illustrate browsing of the Server using a web browser: Figure : The starting page for exploring the Server Browsing a category reveals a list of all the resources of that class, with links to further detail Figure 2: Browsing the list of all eleent sets in the Server When browsing a specific resource, the details fro the RDF description of that resource are displayed, 2003 The Authors 8

9 342 as well as links to related resources Figure 3: Looking at the details of a specific eleent 4.2 The UKOLN Agent Platfor Our ipleentation work has been carried out using the JADE agent platfor. JADE is one of the recoended platfors for developing agent systes. It is a software developent platfor aied at developing ulti-agent systes and applications conforing to FIPA standards for intelligent agents. It includes two ain products, a FIPA-copliant agent platfor and a package to develop Java agents. JADE has provided the environent within which to deploy the ontology service and for building agents. Our platfor has been registered with the platfor directory at Our platfor nae is ukoln.agentcities.net[2]. 4.3 Overview of functionality The Server Agent runs on the UKOLN agent platfor and counicates with the Server using the Server API (over HTTP). It retrieves inforation on eleent sets and returns this inforation in response to requests fro other agents. We have odified the APIs fro the MEG Registry software to support search and browse functions against agency, eleent set, eleent, application profile, eleent usage and encoding schee. Results are returned as RDF-encoded data, rather than HTML. This is possible since the native store of the Server stores the eleent set descriptions as RDF, and uses the Redland RDF toolkit within the HTTP APIs The Authors 9

10 Other Agent Other Agent Platfor Agent Platfor Server Agent Other Agent Server API Ontology Repository Fig 4. Deployent of the ontology server software The Server Agent and two exaples of requester agents are now described The Server Agent The Server Agent can carry out search and browse requests on behalf of other agents, and passes on the results fro the Server to the requester agents. Search Searches are carried out within a specific category (e.g. agency or eleents) and the search ter is atched with any part of the text between the RDF tags aking up a description. If a part of the description atches, the whole description for that resource is returned in the result set. When the description is that of an eleent, the description of the associated eleent set is also presented. Browse Using the browse function, either a whole category is explored, or a specifically naed resource fro a category is specified. The RDF descriptions for all the resources in a category, or for a single resource are returned respectively. Exaples of the RDF (returned in response to both of these kinds of queries) are illustrated in the following sections. Ipleentation Behaviours The Server Agent is ipleented using one behaviour. This behaviour is cyclic and will wait for a essage with a REQUEST perforative. On receiving such a essage, the behaviour. extracts coponents of the request (using an ontology) 2. constructs a URL fro the request 3. connects to the Server using the URL 4. reads the response fro the Server 5. places response into a reply essage Basic error checking is perfored. Incorrect content or an unexpected perforative will result in a NOT_UNDERSTOOD essage being returned to the sender. At present, other error conditions are siply caught within the Java exception echanis and reported on the Syste.err strea. Thus the behaviour deals with one request at a tie, sending a reply before attending to the next request essage in the agent queue The Authors 0

11 A ore coplex odel of behaviour, for exaple starting a new agent or behaviour to deal with each request, was unnecessary at this stage, given the siple functionality of the Server and the agent. In a service level Server, the issue of how to deal with a large nuber of requests in a responsive anner would becoe iportant. The perforance of a large Server capable of coplex querying would also have to be taken into account, but to date such registries are largely an unknown factor Server Ontology We have defined a siple ontology (ServerSearchOntology) in which requests to the Server Agent can be expressed. This ontology is intended to encapsulate the siple kinds of requests supported by the Server that we have experiented with, and is not intended to be an exhaustive or coprehensive ontology for all the kinds of queries that schea registries should or could support The ontology consists of two Action concepts, ReturnSearchResults and ReturnBrowseResults. The ReturnSearchResults action eulates a search request through a web browser; ReturnSearchResults has a searchrequest, ade up of a Scope and a searchter. The scope liits the search for the searchter (which is a string) to one of the categories (agency etc.). ReturnBrowseResults eulates the browsing action carried out through the web browser. Thus a browserequest takes a Scope (one of agency, eleent set, eleent, application profile, eleent usage and encoding schea) and a specific resource URI. The resource URI identifies a specific instance of the entity (e.g. a particular agency) and if a specific resource URI is specified in the browse request, the RDF description for that resource alone is returned. If no resource URI is specified, the RDF descriptions of all the instances of that category are returned in a list (e.g. all the agencies are listed). The exaples illustrate this behaviour Exaples 430 Exaple : An encoding of a search request for the ter "network" within the scope 43 "agency": ( 434 (action 435 (agent-identifier :nae UKOLNServer@solo.ukoln.ac.uk:099/JADE) 436 (ReturnSearchResults 437 (Search :Scope agency :SearchTer network) 438 ) 439 ) 440 ) The RDF description of an agency with the ter resource in its nae is returned: <rdf:description rdf:about=" 445 <rdf:type 446 rdf:resource=" 447 > 448 <reg:agencynae>resource Discovery Network</reg:agencyNae> 449 <reg:agencyhoepage rdf:resource=" 450 </rdf:description> Exaple 2: A search for the ter "audience" in the eleent category ( (action 455 (agent-identifier :nae UKOLNServer@solo.ukoln.ac.uk:099/JADE) 456 (ReturnSearchResults 457 (Search : Scope eleent : SearchTer audience) 458 ) 459 ) 460 ) This search finds two eleents. In the first eleent the search ter 'audience' is found within the 463 usecoent tag. The second eleent is the Audience eleent in the Dublin Core (The search ter is 2003 The Authors

12 highlighted here for ephasis). Both these eleents are part of the Dublin Core Ters eleent set and the description for the eleent set is returned at the end. <rdf:description rdf:about=" <rdf:type rdf:resource=" <rdfs:label>mediator</rdfs:label> <rdfs:coent>a class of entity that ediates access to the resource and for who the resource is intended or useful.</rdfs:coent> <reg:usecoent>the audience for a resource in the education/training doain are of two basic classes: () an ultiate beneficiary of the resource (usually a student or trainee), and (2) frequently, an entity that ediates access to the resource (usually a teacher or trainer). The ediator eleent refineent represents the second of these two classes.</reg:usecoent> <rdfs:subpropertyof rdf:resource=" <reg:iseleentof rdf:resource=" et/dcters"/> </rdf:description> <rdf:description rdf:about=" <rdf:type rdf:resource=" <rdfs:label>audience</rdfs:label> <rdfs:coent>a class of entity for who the resource is intended or useful.</rdfs:coent> <reg:usecoent>a class of entity ay be deterined by the creator or the publisher or by a third party.</reg:usecoent> <reg:iseleentof rdf:resource=" et/dcters"/> </rdf:description> <rdf:description rdf:about=" dcters"> <rdf:type rdf:resource=" et"/> <dc:title>the Dublin Core Ters Eleent Set</dc:title> <dcters:created> </dcters:created> <reg:status>dcmi recoendation</reg:status> <dc:description> The Dublin Core etadata vocabulary is a siple vocabulary intended to facilitate discovery of resources. </dc:description> <reg:responsibleagency rdf:resource=" ci"/> <reg:xlnaespaceprefix>dcters:</reg:xlnaespaceprefix> <reg:specification rdf:resource=" </rdf:description> Exaple 3:A browse request for the whole of the agency category (no Resource URI is given) ( (action (agent-identifier :nae UKOLNServer@solo.ukoln.ac.uk:099/JADE) (ReturnBrowseResults (Browse :Scope agency :Resource "") 2003 The Authors 2

13 ) ) ) Returns a list of all the agencies (descriptions encoded in RDF) <rdf:description rdf:about=" > <rdf:type rdf:resource=" > <reg:agencynae>international Standards Organisation</reg:agencyNae> </rdf:description> <rdf:description rdf:about=" <rdf:type rdf:resource=" > <reg:agencynae>library of Congress</reg:agencyNae> </rdf:description> <rdf:description rdf:about=" > <rdf:type rdf:resource=" > <reg:agencynae>metadata for Education Group</reg:agencyNae> <reg:agencyhoepage rdf:resource=" </rdf:description> <rdf:description rdf:about=" "> <rdf:type rdf:resource=" > <reg:agencynae>oclc</reg:agencynae> </rdf:description> <rdf:description rdf:about=" <rdf:type rdf:resource=" > <reg:agencynae>resource Discovery Network</reg:agencyNae> <reg:agencyhoepage rdf:resource=" </rdf:description> (elided) <rdf:description rdf:about=" "> <rdf:type rdf:resource=" > <reg:agencynae>the Dublin Core Metadata Initiative</reg:agencyNae> <reg:agencyhoepage rdf:resource=" </rdf:description> <rdf:description rdf:about=" The Authors 3

14 <rdf:type rdf:resource=" > <reg:agencynae>world Wide Web Consortiu</reg:agencyNae> </rdf:description> Exaple 4: A browse request for a specific resource ( fro the encoding schee category. ( (action ) ) (agent-identifier :nae UKOLNServer@solo.ukoln.ac.uk:099/JADE) (ReturnBrowseResults (Browse :Scope encodingschee :Resource ) 4.4 Interrogating the Server Agent We have ipleented two exaples of Requester Agents, both of which are driven by a huan user and ake requests to the Server Agent. These two agents use the ServerSearchOntology to counicate requests to the Server Agent, and display the response returned by the Server. Results to queries are contained within the content slot of an INFORM essage fro the Server Agent, and consist of RDFS descriptions. Thus the ontology is only used to counicate requests; responses are siply wrapped up in the content slot of the essage The GUI Agent This agent presents the user with a graphical interface ipleented with Java Swing. This is realized through two classes: ServerAgentGui class extends the Swing JFrae class, and defines the appearance of the interface; ServerGuiAgent class extends the Jade GuiAgent class, and defines the behaviours that are instantiated in response to user actions at the interface. Each instance of the Agent class is associated with one instance of the Gui class (and vice versa). The appearance of the interface is shown in Figure 5. It contains the following ain coponents: a pull-down list of categories a button for triggering the display of a whole category (the latter obscured in the first screenshot) a text entry for resource URIs, and an associated button for displaying; a text entry for search ters, with an associated search button a display area for results After selecting a category, the user can then choose to browse the whole category, or to enter a resource URI for a known resource. Alternatively, the search box can be used to interrogate the Server. The three tasks that the interface supports reflect the kinds of requests that can be expressed in the ServerSearchOntlogy: The ServerAgentGui class ipleents ActionListener; on Action events, the handler (ActionPerfored) invokes the JADE postguievent ethod to counicate with the ServerGuiAgent class; this is the path by which user actions on the interface trigger behaviours in the agent. Within the agent, the onguievent ethod handles the events fro the interface (invoked through postguievent). A essage is built (using the ServerSearchOntology) corresponding to the action invoked; the essage is sent using a SenderBehaviour (which extends OneShotBehaviour). A cyclic behaviour listens for response essages fro the ServerAgent 2003 The Authors 4

15 and when an INFORM essage arrives, it invokes a displayresults ethod in the gui, so that the content of the essage (containing RDF-encoded descriptions) is displayed (Figure 6). The interface has been design to support one outstanding request at a tie. In theory ultiple requests could be launched before the first response arrives, and at present there is no control to prevent this. In practice the syste response is sufficiently fast that no ajor control is required at present to synchronise requests and responses. If such control were required, this could best be ipleented through the Gui by disabling the sending controls until a response is received. An alternative would be an interface that supported ultiple outstanding requests, but this would require a ore coplicated design that is beyond the scope of the present project. This also requires a ore coplicated coordination odel between the interface and the agent(s) for anaging requests. The link between the ServerGUIAgent and the Server Agent is hardwired and the Server Agent is assued to be running locally Figure 5. Using the interactive GUI of the ServerGuiAgent to enter requests 2003 The Authors 5

16 Figure 6. Results are displayed in a window in the GUI The Coand Line agent A second Agent Class, ServerRequesterAgent, has been provided to interact with the user through the coand line. On setup() this agent first establishes which Server the user would like to use, with a choice of either the UKOLN Server, or a local one Behaviours The Agent then instantiates a ain sequential behaviour (HandleRequestsBehaviour) which propts for and reads input fro the terinal. The onstart() ethod of the ain behaviour interacts with the user to define what kind of transaction the user is perforing (browse or search) and its paraeters: scope, search ter or resource URI: ENTER the local nae of the Server agent or press enter to use the UKOLN Server--> ENTER s for search or b for browse --> s Class to Search ---> eleent Enter a SearchTer ---> audience A suitable essage is then built and a Sender Behaviour is scheduled (as a sub behaviour) to send the essage to the Server Agent. The next subbehaviour added then handles the response fro the Server Agent and displays the result to the user. The onend() ethod then checks if the user would like to carry out another transaction. If the user stops, the agent is terinated; if the user wishes to continue, all the behaviours are reset The Authors 6

17 Conclusions We have successfully deployed an ontology server onto the Agentcities.NET network, where it is available for either browsing over the Web or querying by agents. It should be noted that the server accepts etadata vocabularies encoded in RDF Schea. Further, the vocabularies need to adhere to the odel described in the Appendix. The work presented has advanced the work begun in previous projects to investigate an approach based on autoated querying and processing of siple ontologies by software agents rather than through huan interaction. 6 Acknowledgeents The software used in this project for the ontology server was originally developed in the MEG Registry project which was funded by JISC and BECTa. The ideas in this project have been developed fro work in the DESIRE, SCHEMAS and MEG Registry Projects. Thanks to Pete Johnston, UKOLN, University of Bath, for help with the MEG Registry and associated software. Thanks also to Owen Cliff, Departent of Coputer Science, University of Bath for assistance with setting up the UKOLN server. UKOLN is funded by Resource: The Council for Museus, Archives & Libraries, the Joint Inforation Systes Coittee (JISC) of the Higher and Further Education Funding Councils, as well as by project funding fro the JISC and the European Union. UKOLN also receives support fro the University of Bath where it is based The Authors 7

18 7 References [] UKOLN web-page for Agentcities.NET deployent project [2] UKOLN Ontology Server for the Agentcities.NET network [3] UKOLN [4] Agentcities.NET [5] Dublin Core Metadata Initiative (DCMI) [6] DESIRE [7] SCHEMAS Project [8] Thoas Baker, Makx Dekkers, Rachel Heery, Manjula Patel, Gauri Salokhe, What Ters Does Your Metadata Use? Application Proifles as Machine Understandable Narratives, Journal of Digital Inforation Vol 2 (2), Noveber [9] Heery H and Patel M., Application Profiles: ixing and atching etadata scheas, Ariadne Issue 25, Septeber [0] OntoWeb Technical Roadap v [] Nubered Hypernotes in J.Hendler ref&siteid=sci ref&siteid=sci#note9 [2] Sowa, J.F. Building, Sharing and Merging Onotologies: Glossary [3] SCHEMAS Project Glossary [4] MEG Website [5] MEG Registry Project [6] Beckett D., The Design and Ipleentation of the Redland RDF Application Fraework Proceedings of WWW0, Hong Kong, May [7] The MEG Registry and SCART: copleentary tools for creation, discovery and re-use of etadata scheas. Rachel Heery, Pete Johnston, Dave Beckett (ILRT, University of 2003 The Authors 8

19 [27] WordNet Bristol) & Daian Steer (ILRT, University of Bristol) - October 2002 In: Proceedings of the International Conference on Dublin Core and Metadata for e- Counities, Florence: Firenze University Press, 2002, pp [8] RDF Scheas [Note: this is the latest version of the W3C Working Draft, released on 2 Noveber 2002, which is a work in progress; The registry developent took place before the release of this draft.] [9] W3C Web Ontology Working Group [20] SWAD-Europe: Scalability and Storage: Survey of Free Software /Open Source RDF storage systes [2] DAML+OIL [22] OWL [23] OWL2 [24] Wilson, Michael [25] SeanticWeb.org [26] Hendler Web version of IEEE article [28] RDFS(FA) [29] The Meg Registry Client Software (SCART) [30] DAML Repository [3] BT Ontology Server [32] BT Ontology Server Service Description [33] Another Ontology Page [34] SCHEMAS: Best practice guidelines for anaging a registry The Authors 9

20 835 [35] DCMI Registry Working Group [36] OWL Overview [37] EOR (Extensible Open RDF) Toolkit The Authors 20

21 Appendix: The MEG Registry Data Model Eleent Set Agency Eleent Encoding Schee Value Eleent Usage App Profile Agency: An organisation or individual responsible for anaging one or ore Eleent Sets, Application Profiles or Encoding Schees Relationships Eleent Set is-managed-by (-) Agency Encoding Schee is-managed-by (-) Agency Application Profile is-managed-by (-) Agency Agency Properties Identifier (URI) Nae Hoe Page URL The nae or title of the Agency A source of further info about the Agency Eleent Set: A set of etadata Eleents that is anaged as a coherent unit by an Agency. The Eleents of an Eleent Set are functionally related, by virtue of having been defined for the purpose of usefully describing the characteristics of a resource Relationships Eleent Set is-managed-by (-) Agency Eleent is-eleent-of (-) Eleent Set Eleent Set Properties Identifier (URI) Title The nae or title of the Eleent Set 2003 The Authors 2

22 Version Date created Status Description Classification Specification The version of the Eleent St Date this version created Draft/recoendation etc Including any notes of scope/purpose Prose description of/guidelines for use of Eleent Set Eleent: A forally defined ter that is used to describe a characteristic or attribute of a resource Relationships Eleent is-eleent-of (-) Eleent Set Eleent associated-encoding-schee (-) Encoding Schee Eleent refines (-) Eleent Eleent Usage uses (-) Eleent Eleent Properties Identifier (URI) Nae Definition Coent Data type Obligation Maxiu occurrence A huan-readable version of the property nae A stateent that clearly represents the concept and essential nature of the Eleent A reark concerning the application/use of the data eleent Indicates the type of data that can be represented in the value of the data eleent Indicates whether the Eleent is always or soeties required to be present Indicates any liit to the repeatability of the Eleent Encoding Schee: A set of contextual inforation or parsing rules that aids in the interpretation of the value of a etadata Eleent. Encoding Schees include controlled vocabularies, which enuerate a list of values, and; foral notations or parsing rules, which define precisely how a lexical representation of a value is to be interpreted Relationships Encoding Schee is-managed-by Agency (-) Agency Eleent associated-encoding-schee (-) Encoding Schee Eleent Usage associated-encoding-schee (-) Encoding Schee Value type (-) Encoding Schee Encoding Schee Properties Identifier (URI) Nae Version Date created Status Description Classification Specification The nae or title of the Encoding Schee The version of the Encoding Schee Date this version created Draft/recoendation etc Including any notes of scope/purpose Prose description of/guidelines for use of Encoding Schee Controlled Vocabulary Value: An individual value or ter in a controlled vocabulary Relationships Value type (-) Encoding Schee 2003 The Authors 22

23 Identifier (URI) Value Label Description Value Huan-readable for of value Explanation or definition of value Application Profile: A set of Eleent Usages that is anaged as a coherent unit by an Agency. An Application Profile is optiised for the resource description requireents of a particular application or context. Like the Eleents of an Eleent Set, the Eleent Usages within an Application Profile are functionally related, by virtue of having been defined for the purpose of usefully describing a resource. Within an Application Profile, the Eleent Usages ay reference Eleents fro ultiple Eleent Sets Relationships Application Profile is-managed-by Agency (-) Agency Eleent Usage is-usage-in (-) Application Profile Application Profile Properties Identifier (URI) Title Version Date created Status Description Classification Associated XML Schea Specification The nae or title of the Application Profile The version of the Application Profile Date this version created Draft/recoendation etc Including any notes of scope/purpose Prose description of/guidelines for use of Application Profile Eleent Usage: A deployent of a (previously defined) etadata Eleent in the context of a particular doain or application. The used Eleent ay be tailored for the context by: a narrowing of its seantic definition; association with specified datatypes or Encoding Schees; specification of obligation/occurrence constraints Relationships Eleent Usage is-usage-in (-) Application Profile Eleent Usage uses (-) Eleent Eleent Usage associated-encoding-schee (-) Encoding Schee Eleent Usage Properties Identifier (URI) Nae Definition Coent Data type Obligation Maxiu occurrence A huan-readable version of the Eleent nae. A stateent that clearly represents the concept and essential nature of the Eleent A reark concerning the application/use of the Eleent. Indicates the type of data that can be represented in the value of the Eleent Indicates whether the Eleent is always or soeties required to be present Indicates any liit to the repeatability of the Eleent 2003 The Authors 23

24 The Authors 24