Ontology for Coordinating Dialogs in Distance Learning Environments Marco A. Eleuterio Jean-Paul Barthès PUC-PR Université de Technologie de Compiègne Rua Imaculada Conceição, 1155 Rue Personne de Roberval Curitiba – PR - Brasil BP 20529 - 60205 Compiègne Cedex (55)-41-330-1555 (33) 03 44 20 48 13 marcoa@ppgia.pucpr.br barthes@utc.fr ABSTRACT worksheets) or finishing the discussion. This process is This paper discusses the role of ontology for an intelligent agent continued until a satisfactory degree of collective agreement is designed for coordinating domain dialogs among participants of a achieved. The discussion is organized as an argumentation tree distance learning environment. The domain models used by the [1], where each node corresponds to a dialog argument. Most of agent are centered on the ontological representation of the the intelligent behavior featured by the system is the result of (i) domain concepts. The agent’s behavior as well as the adequacy the regrouping algorithm that takes into account structural and of the domain models have been tested in actual distance semantic parameters and (ii) the dynamic generation of learning situations. discussion elements, both services being delivered by the KB agent. 1. INTRODUCTION Distance learning environments are widely used to allow full 3. THE KB AGENT AND THE ROLES OF distance courses or used as a complement to more traditional THE ONTOLOGY classes. Due to their inherent distributed nature, such The role of the KB agent is thus to perform all domain-specific environments have a large potential for agent architecture, either tasks, i.e. generate discussion elements, perform semantic for delivering learning material or enabling communication matching and evaluate the semantic coverage of the discussion. among the participants in so-called discussion forums. In either The domain ontology, in our work, is the central representation aspects, the ontological representation of the domain plays an that provides the KB agent with the required amount of theory- important role. We are developing an agent system capable of awareness. coordinating collective discussions in distance learning environments by using a set of specialized agents. Two of the 3.1 Ontology for generating discussion major agents of this system are the dialog agent and the elements knowledge base (KB) agent. The generation of text-based questions in natural language (or content-expected interrogative speech acts [2]) starts up a The dialog agent coordinates the discussion by generating series discussion tree. We represent the domain by using two different of dialog cycles and maintaining an argumentation tree. The KB models: the domain ontology and the task structure. The agent performs all domain-related tasks, being the ontology its ontology relates domain concepts by means of part-of and is-a central knowledge model. Considering the purposes of this links, and is used to represent the concepts manipulated by the workshop, we will give more emphasis to the KB agent. tasks of the task structure. The ontology is used to produce questions like: what kinds of do you identify, or what 2. THE DIALOG AGENT AND THE are the composing elements of ? Such questions are ARGUMENTATION STRUCTURE produced so as to cover a certain number of concepts scheduled In this section we briefly describe the dialog agent’s behavior. for the discussion. This agent has the task of initiating, coordinating and closing the discussions by generating dialog cycles. Initially, given a set of 3.2 Ontology for semantic matching When a discussion element is generated, a central concept is questions (discussion elements, or DEs) and a set of participants, identified. It is the one appearing in the text. All related sub- the dialog agent builds and distributes a set of questions to the concepts that appear in the domain ontology are considered to be participants. As soon as the questions are answered, the agent the sub-domains of this discussion element. By analyzing the reshuffles the groups of participants and sends the questions with occurrence of such concepts in other answers or comments, the their answers to be analyzed and commented by the new groups. agent can discover semantic relations and use them for further According to the level of agreement and the content of the regrouping of the participants for the next discussion cycle. comments, the system may decide upon triggering another discussion cycle (regrouping the participants and building new Workshop on Ontologies in Agent Systems (OAS2001) - Autonomous Agents 2001, may 2001, Montreal, Canada. 4. OUR EXPERIENCE IN DOMAIN ontology by means of specialized links, namely output resource, input resource and implicit knowledge resource, which specify MODELING how a certain concept is used by the task (see Figure 1). We have conducted experiences in domain modeling as part of a research project between the Technology University of Compiègne (UTC) and CEGOS, a French enterprise that 4.3 Implementation issues provides on-line training. We chose a specific CEGOS course We implemented both the ontology and the task model in LISP, and built the knowledge models for it, i.e., the task model and as two independent structures linked together by a set of the domain ontology. We are now designing the agents’ specialized links. mechanisms based on such models. The items below describe the major results from this project concerning the KB agent. We also implemented an editor that allows the domain experts of CEGOS to build and edit their own ontologies (see Figure 2). 4.1 A “what-for” approach for designing the ontology Given the problem, we started with analyzing what the ontology would be used for, and then we chose a representation for it. For the purpose of generating discussion elements the ontology, as well as the task model, should provide elements for building interrogative sentences. Such sentences are meant to investigate the domain along five different axes: (i) the nature of the concepts (ontology is-a links); (ii) the elements of a composed concept (ontology part-of links); (iii) the use of the concepts by a certain task (task models resource link); (iv) the decomposition of a complex task into sub-tasks (task model sequence link); and (v) different ways of performing a task (task model type link). The second use of the ontology is to perform semantic matching, a process by which the agent dynamically regroups the participants of the discussion according to the content of their answers or comments. To this purpose, mainly is-a and part-of links are used to measure the semantic distance between two text Figure 2: Ontology and Task Model Editor chunks. 4.2 Domain modeling The course we modeled, named “Le responsable formation 5. CONCLUSION nouveau dans sa fonction”, which can be roughly translated as In our work, we identified the need for an ontology and looked “How to manage competence in an enterprise” covers several for an adequate representation for it. The complexity of an different domains, ranging from human resource administration, ontology, however, is related to the type of use we intend for it. to teaching methodologies and legal aspects. The diverse nature Our problem requires a terminological ontology, i.e., a of the course content lead us to organize the needed ontology as a structured collection of terms. Other applications may need more collection of domains. powerful ontologies, like those containing formal definitions (interpretable ontologies), or even executable ontologies based The strong “how-to-do” feature of the content lead us to make on the notion of task ontology and abstract code [3]. use of another model, the task structure, that represents the tasks of a training a manager in her daily work. The next obvious 6. REFERENCES choice was to link the task structure to the corresponding domain [1] Karacapilidis, N; Papadias D. A computational approach for argumentative discourse in multi-agent decision making Ontology environments. AI communications 11 (1998) 21-23. Task model (a set of domains) [2] Porayska-Pompa, K; Pain, H. Aspects of Speech Act Output resource Categorisation: Towards Generating Teacher’s Language. Input resource International Journal of Artificial Intelligence in Education (2000). Implicit knowledge Is-a-link Is-a-link Seq-link Is-a-link Part-of-link [3] Mizoguchi, R; Bourdeau, J. Using Ontological Engineering Type-link Part-of-link Part-of-link to Overcome Common AI-ED Problems. International Journal of Artificial Intelligence in Education (2000). Figure 1: Link between the task model and the domain ontology