- From: Leo Obrst <lobrst@mitre.org>
- Date: Thu, 13 Dec 2001 10:59:49 -0500
- To: W3C Web Ontology WG <www-webont-wg@w3.org>
W3C Web Ontology WG: Content Interoperability Use Case December 13, 2001 The following use case includes descriptions and examples gleaned from the general mail regarding use cases, but focused purely on the Content Interoperability Use Case. It also includes points from the rather limited discussion our use case group has had up to this point. It is still very sketchy and should not yet be considered a consensus view. I've used Guus Schreiber's general format for the use case. Participants: Leo Obrst lobrst@mitre.org Pat Hayes phayes@ai.uwf.edu Dan Connolly connolly@w3.org Mike Dean mdean@bbn.com Herman ter Horst herman.ter.horst@philips.com Deborah McGuinness dlm@ksl.stanford.edu Raphael Volz volz@fzi.de --------------------------- SCOPE AND DEFINITION --------------------------- General Description: Content Interoperability includes notions such as Semantic Interoperability and interoperability issues dealing with the distribution and mapping of content to devices. Semantic Interoperability is roughly the capability to send and receive content supported by ontologies across applications with retention of semantics, the ability to map between different ontologies in a semantics preserving manner. Other aspects: semantic equivalence, determining the consistencies between two ontologies, ontology integration and merging. Included under this use case are: - semantic mapping from database schemas to ontologies - semantic mapping from taxonomies and classification systems to ontologies - semantic mapping from ontologies to ontologies - ontology partitioning and decomposition as the flip of ontology integration and ontology merging - ontology language extensions to support mapping ontologies - electronic commerce standards and catalog mappings using ontologies - dynamic composition of web services supported by independent ontologies - construction of an ontology "view" or presentation based on the needs of a user, application, specific task using an ontology or ontologies, and a device - translating between ontologies Content Interoperability Subtasks: - conceptual search to support ontology mapping - versioning issues: once mappings are created, are they saved, are they updated, what about when the ontologies themselves change over time - support in the ontology language for user, task, application view and context descriptions, constructions --------------------------- USE CASE DESCRIPTIONS (all are excerpted from email messages) 1. CONTRIBUTOR: Mike Dean TASK: Automate tasks relating to travel requiring interacting ontologies and usable by palm pilot device DESCRIPTION: I hope that WebOnt will be used to provide information that's currently available on the WWW (and not currently available on the WWW) in such a way that I can write and/or use programs to automate tasks such as those related to business travel. I'll use that domain as a focus for this use case. I generally plan my travel before calling my human corporate travel agent. For flights, I use a copy of the United Airlines Electronic Timetable, which gets monthly (weekly since 9/11) updates in a some sort of compressed binary format (I've made several unsuccessful attempts to extract the underlying data). I like the fact that I can work with this offline (e.g. on an airplane). I'd much prefer to get it in WebOnt format so that I can apply my own preferences, link to other information about airports, etc. I have a hotel chain that I prefer. For destinations that I frequent often, I know their hotels in the area but often forget some of the details (e.g. which ones serve good hamburgers, and which room locations to avoid). For others, I look up information on their web sites. I'd like to get this information in WebOnt format, and to add my own properties for items of personal interest. I normally try to get a hotel room that has high-speed wired or wireless Internet access. Hotel web sites are very inconsistent in reporting this service. For an unfamiliar property, I generally look at the directories maintained by the major ISPs serving hotels (CAIS, STSN, Wayport, and MobileStar), none of which currently provide the information in an agent-friendly format (most use maps, page hierarchies, and/or PDF). I'd prefer to get it in WebOnt and merge it myself with my itinerary and other geographic information. When I make a reservation, my corporate travel agent emails my itinerary in a format that I consider a canonical example of the "un-Semantic Web": a PDF image of a traditional FAXed itinerary. This prints well, but is virtually impossible for a program to process. I'd prefer to get this content using WebOnt, and to have it automatically routed to a personal travel agent program. I'd like to automatically share some of my itinerary information (e.g. travel dates and arrival times) with my co-workers, but keep some of it (e.g. credit card numbers) private. After my airline ticket is booked, I generally have to call the airline directly to get an upgrade and/or better seat assignment. I prefer non-bulkhead (so I can keep my laptop under the seat in front of me) window seats. United is unable to track this preference, so I often have to make several calls. An agent would be better (and more reliable!) at this task. I now subscribe to United's Flight Paging Service, which automatically sends an email message to my pager 2 hours before my flight or whenever a delay or cancellation occurs. I'd prefer for my agent to get this information in WebOnt format so that it could automatically begin identifying alternative flights and routings when a problem arises. I also subscribe to a free service from fly.faa.gov, which sends me email messages on ground stops and delays at specified airports. Unfortunately, it's not linked to my itineraries, so I get lots of such messages while I'm not travelling. If the information was in WebOnt format, my agent could easily cross-reference it with my itineraries and identify relevant problems. While I travel, I'd like to have fast access to my itinerary using a utility like PalmDAML [1] on my PDA. When I have a substantial wait at an airport, I like to look for high-speed Internet access. I've generally had better luck searching concourses than web pages to find such services; I'd like to get such information (translated to) my preferred WebOnt ontology. I sometimes go to the American Airlines Admirals Clubs to use their high-speed MobileStar wireless Internet access points. This is usually in a different terminal, so I'd like to also get WebOnt information on gate locations and walking times. When I go to an unfamiliar city, I often try to rent a Hertz car with a NeverLost GPS. Rather than painstakingly toggling in the street names and numbers for my hotel and other destinations, I'd like to just beam the information in WebOnt format from my PDA using IR or Bluetooth. I'd also like to get additional information that's not generally now on the web: service hours for restaurants (and room service) along my travel route. For flights that get in late, for example, my agent could tell me if I need to grab a bite before leaving the airport. When I return from a trip, I have to fill-out an Excel spreadsheet for my expense report. Most of this information could come directly from my itinerary, hotel bills, and credit card receipts if they were provided in WebOnt format. I already have a DAML application [2] for reconciling my expense reports with my credit card statements and checking account. A few observations: 1) most of this information (flight schedules, travel itineraries, hotel addresses, expense reports, etc.) is not ontologically sophisticated 2) much of the information is already available in human-readable form 3) automation currently exists only in specific stovepipes such as United's new Flight Paging Service 4) even a highly-motivated geek finds it impractical to merge the existing information 5) with widespread use of WebOnt, we should be able to do most of these things pretty easily [1] http://www.daml.org/PalmDAML/ [2] http://www.daml.org/2001/06/expenses/ EXAMPLE DOMAIN: Travel TYPICAL USER: traveler needing to plan and schedule travel accommodations possibly from multiple devices WEB ONTOLOGY REQUIREMENTS: - correlation/mapping/merging of ontologies representing diverse information sources and services - distribution of ontology information to different devices, hence issue of condensation to low bandwidth, etc. ===== 2. CONTRIBUTOR: Jonathan Dale TASK: Ontology representation exchange and translation DESCRIPTION: Both FIPA and Agentcities are aiming towards the practical application of agents and agent technologies, so they are looking at choosing an ontology representation language (ORL) from a pragmatic standardisation perspective. - To assist in ontology representation exchange. Initially, we expect that FIPA and Agentcities will develop ontologies in a human-centric, collaborative manner since most application domains that they are trying to define are reasonably small and finite (i.e., bottom up rather than top-down). However, in the future, it will be important that an ORL can also express large ontologies and can reference terms in other ontologies, possibly in other ORLs. - To assist in ontology translation. As the nodes in the network of FIPA-compliant agent platforms increases, so the heterogeneity in the network increases. FIPA is based on a model of uinting heterogeneity through interoperability, and a key feature of a suitable ORL should be like-compatibility with other ORLs. If the functionality is too diverse, then translation between ORLs will be more difficult. EXAMPLE DOMAIN: Intelligent agent communities TYPICAL USER: end user needing services, intelligent agent developer WEB ONTOLOGY REQUIREMENTS: - need to express large ontologies referring to terms in other ontologies - need to translate between ontologies ===== 3. CONTRIBUTOR: Ned Smith TASK: Management of control networks and their integration into corporate LAN or other types of networks using ontologies for different devices and composite devices DESCRIPTION: Use Cases: 1) A sensor device manufacturer builds a device that interoperates with sensor/actuator devices built by other manufacturers. The device properties are expressed in ontological form. The ontology of device properties or "device ontology" is embedded in the device. If the device is connected to a network, it can be recognized by and installed into that network by an agent that parses the device ontology and determines how best to integrate its function. Once installed, the device may be bound to other devices forming a composite device. The composite device, logically a unique device, may interact with other devices or services on the network. Device manufacturers are not expected to perform a'priori testing of the possible device configurations to achieve interoperability. 2) A legacy control network performs a process that administrators do not want to disturb. However, they do want to monitor certain functions. They build an ontology of the control network / process and map monitored functions into properties of the ontology. Outside services may discover the control network ontology. Property value changes can trigger notification events sent to outside services. The monitoring functions may NOT introduce delays or in any way prevent the control network from performing its task. The monitoring subsystem may be re-configured by administers periodically and will not impact the monitored control process. 3) A control network is managed by multiple outsourced management service providers. Management responsibilities are delegated by the control network owner to the service providers. Management responsibilities are divided among the service providers in such a way as to prevent administration overlap. Management actions are verified to be acceptable prior to their application. Service providers may re-negotiate how responsibilities are divided periodically. After which previously granted privileges are lost and new privileges granted. EXAMPLE DOMAIN: Control network management TYPICAL USER: Control network management administrators and service providers WEB ONTOLOGY REQUIREMENTS: 1a) The ontology representation language (ORL?) must be simple and concise enough to represent device properties in a small memory space. Devices range in complexity - economies of scale suggests the smaller the space required the better. If the device ontology is unacceptably large, a unique reference to the ontology must be available. 1b) The device ontology must be linkable with other devices' ontology forming a distributed ontology with a logical beginning and end / deterministic traversal method. 1c) Discrete partitions of an ontology must be recognizable in terms of the physical boundaries and in terms of logical (composite device) boundaries. 1d) The device is aware of its own ontological properties. 1e) If a composite device has properties in excess of the sum of the properties of subordinate devices, these properties can be physically located on multiple nodes. 1f) Device properties may be owned/controlled by multiple entities. The owner/controller may modify property values. 1g) Both property type and value can be changed. Type changes can be discovered by other services/devices and type semantics can be resolved - hopefully quickly! 2a) Properties in an ontology must be associate-able with a service/function, that supplies the property value. 2b) Properties must support access preconditions - namely, the monitoring network should be prevented from writing, while allowing write by the control network. 2c) Property values have meta attributes describing data freshness, polling/event interval or freshness goal, type, acceptable value range(s), exception conditions. 2d) Administrative update are access controlled and versioned allowing only authorized updates and rollback should an update not work as intended. 3a) Resources in the control network are described by an ontology. 3b) Resource can be managed by different / multiple entities. 3c) Management responsibility may be delegated. 3d) Managed resources have meta attributes describing management functions - namely actions, time intervals, audit log, owner/controller/delegate and synchronization events. ===== 4. CONTRIBUTOR: Stefan Decker TASK: Support dynamic aspects of device configuration using ontologies DESCRIPTION: within LastMileServices: ---------------------------------- LastMileServices is a startup aiming to describe static and dynamic aspects of telecommunication devices, with the goal of simplifying service construction and configuration of large networks. We use on ontology language to define device ontologies (e.g., router and switches) For capturing the dynamic aspects of device configuration we have developed a service description ontology, which defines primitives to declare task decomposition, control flow (using the vocabulary of a UML statechart), and data flow of services. Other primitives enable to create new services by combining and configuring existing services. This results in a compositional ontology-based process and service description language capable to combine existing services (given by, e.g., distributed Web Services or Java Objects) with the goal to create new services. A service description can be compiled to Java code and be executed. EXAMPLE DOMAIN: distributed web services, device configuration TYPICAL USER: web service or device user, web service developer, device developer WEB ONTOLOGY REQUIREMENTS: A compositional ontology-based process and service description language capable of combining existing services ===== 5. CONTRIBUTOR: Mike Sintek TASK: Design a scalable, agent-based middleware for distributed Organizational Memories (OM) for knowledge management using ontologies, especially mapping of ontologies across OMs. DESCRIPTION: The following are some passages from [1] which describes the use of ontologies in the FRODO [2] project (and for Knowledge Management in general). In the FRODO project, we design a scalable, agent-based middleware for distributed Organizational Memories (OM). In knowledge management (KM), it is widely accepted that ontologies as explicit specifications of conceptualizations [Gruber91] provide a useful means to facilitate access and reuse of knowledge. Typical utilization scenarios comprise discussion groups, search engines, information filtering, access to non-textual information objects, and expert-user communication. In the context of Organizational Memory (OM), ontologies provide a vocabulary for specifying information resources as well as information needs in order to evolve from a keyword-based towards a concept-based information management, indexing, and retrieval approach. They also form the basis for knowledge-enhanced or knowledge-assisted search and retrieval. In these applications, ontologies serve as formally represented "specifications of discourse in the form of a shared vocabulary". Such a shared understanding is particularly important because KM typically deals with multi-actor scenarios. The vision of knowledge management assumes the comprehensive use of an enterprise's knowledge, whoever acquired it, whereever it is stored and however it is formulated in particular. Technical support for such a vision is often based on *centralized approaches* which seem well-suited to guarantee that the *complete* information available is considered. For instance, in OM reference architecture (see attached figure) derived from the KnowMore framework, the problem of several heterogeneous information sources is tackled by the introduction of a uniform *knowledge description level*: The various information items are annotated by knowledge descriptions which are based on an agreed upon vocabulary, namely the information, enterprise, and domain ontologies. Hence, a centralized view upon a distributed information landscape is built. In the FRODO project we aim at extending the centralized (KnowMore) framework towards a *distributed OM* scenario, resulting in a need for *domain ontology services*. This requires facilities for both adding domain ontologies to an OM and accessing ontology services from other OMs. Therefore we propose two types of ontology services: Domain Ontology Agents (DOA) and Distributed Domain Ontology Agents (D2OA). Domain Ontology Agents are responsible for ontologies *within one OM*, Distributed Domain Ontology Agents are located *between several OMs* and facilitate cross-OM communication. So, the task of D2OAs is quite similar to "standard information integration ontologies" (e.g. mapping services), but much easier as the sources are already formal ontologies, not just "any information provider". Typical questions to DOAs are "What are the subconcepts of concept A?" whereas D2OAs answer questions like "Which OM contains concepts like A and B?" or "What does A mean in OM_y?". This structure better embraces the inherently distributed nature of (ontological) knowledge. Not *all conceptualizations* are shared between *all actors* of the system, but *ontology societies* are formed with respect to relevant domains. Additional infrastructure enables communication between these ontology societies. [1] http://citeseer.nj.nec.com/450299.html [2] http://www.dfki.uni-kl.de/frodo/ EXAMPLE DOMAIN: Knowledge management and in particular, Organizational Memories TYPICAL USER: end user looking for information in an organization or across organizations WEB ONTOLOGY REQUIREMENTS: - support for mapping between ontologies - support for answering about multiple ontologies: does the ontology have concept A?, what does A mean in ontology X? ===== 6. CONTRIBUTOR: Jeff Heflin TASK: Context sensitive and inter-domain search on the Web DESCRIPTION: - context sensitive web search - inter-domain web search - question answering on the Web EXAMPLE DOMAIN: TYPICAL USER: User searching the Web WEB ONTOLOGY REQUIREMENTS: These requirements describe the features needed by complete Web Ontology language. I have developed this list over time as a result of lessons I learned from my experiences with SHOE and DAML+OIL. Some of these requirements may be out of scope of the WebOnt Working Group, and instead should be saved for later "layers" of the Semantic Web. Nevertheless, identifying them now can help us to ensure that design of the Ontology layer does not prohibit the addition of them later. <excerpted:> management of ontology evolution [HH00, KF01, Hef01 (Section 3.4)] version numbers or revises relation semantic backward-compatibility deprecation interoperability of heterogeneous ontologies features for mapping between ontologies synonyms including synonyms for individuals! subclass / superclass inverse relationships conjunction, disjunction, implication arithmetic functions grouping string manipulation procedural attachments? Java? management of inconsistency inconsistency is inevitable in large, distributed systems need standard meaning in face of inconsistency ontologies of trust/belief prioritized conflict handling [GLC99] assumption-based truth maintenance systems (ATMSs) [dK86] non-monotonicity other methods? scalability limit expressivity? expect incomplete reasoning algorithms? References [dK86] J. de Kleer. An Assumption-based TMS. Artificial Intelligence, 28(2), pp. 127-162, 1986. [GLC99] B. Grosof, Y. Labrou, and H. Chan. A Declarative Approach to Business Rules in Contracts: Courteous Logic Programs in XML. In Proc. 1st ACM Conf. on Electronic Commerce (EC-99), 1999. (PDF) [Hef01] J. Heflin. Towards the Semantic Web: Knowledge Representation in a Dynamic, Distributed Environment. Ph.D. Thesis, University of Maryland, College Park. 2001. (PDF) [HH00] J. Heflin and J. Hendler. Dynamic Ontologies on the Web. In: Proceedings of the Seventeenth National Conference on Artificial Intelligence (AAAI-2000). AAAI/MIT Press, Menlo Park, CA, 2000. pp. 443-449. (PDF) [KF01] M. Klein and D. Fensel. Ontology Versioning on the Semantic Web. In First International Semantic Web Working Symposium (SWWS'01), 2001. (PDF) ===== 7. CONTRIBUTOR: Leo Obrst TASK: Fusion/composition of information from disparate ontologies DESCRIPTION: Enable the fusion/composition of information from disparate ontology modeled sources to a resolved or reference model (for defense-related command and control decision-making, for example). EXAMPLE DOMAIN: Defense command and control TYPICAL USER: Command and control analyst WEB ONTOLOGY REQUIREMENTS: - capability of composing information supported by diverse ontologies -- _____________________________________________ Dr. Leo Obrst The MITRE Corporation mailto:lobrst@mitre.org Intelligent Information Management/Exploitation Voice: 703-883-6770 7515 Colshire Drive, M/S W640 Fax: 703-883-1379 McLean, VA 22102-7508, USA
Received on Thursday, 13 December 2001 11:00:20 UTC