OWL-OLM: OWL-Based Dialogue for User Knowledge Acquisition

User Modelling and User Adaptive Systems Activity
Knowledge Representation and Reasoning Research Group
School of Computing, University of Leeds

People: Ronald Denaux (Technical University of Eindhoven), Lora Aroyo (Technical University of Eindhoven),
Vania Dimitrova (University of Leeds), Michel Pye (University of Leeds)

Funding: British Council and NWO (UK-NL Partnership in Science) and PROLEARN EU NoE

Motivation: User Knowledge Acquisition

A critical aspect in the realisation of personalisation is the acquisition of knowledge about the users. Traditionally it represented in a user model. The depth of the represented understanding about the users can vary from simple user profiles that focus mainly on users' preferences to sophisticated models that capture users' conceptualisations. In the open world context the later capture individual users' viewpoints, spanning from knowledge engineers to naive users. They also define the semantics of the user knowledge representation, thus facilitating the effective integration of the users within semantics-aware systems. Furthermore, this enables knowledge-enhanced reasoning to align the perspectives of users and system designers, where various mismatches in meaning and expression can be discovered and taken into account for more effective adaptation.

The acquisition and maintenance of user conceptual models requires robust methods that integrate seamlessly in semantics-enhanced Web-based systems. OWL-OLM is an example of such an approach which has been demonstrated in an e-learning context.

Description of OWL-OLM

OWL-OLM is an OWL-based framework for Open Learner Modeling to: OWL-OLM follows the STyLE-OLM framework for interactive open learner modeling but amends it to work with a domain ontology and user model built in OWL.
OWL-OLM Architecture.
The architecture of OWL-OLM.
The Dialogue Agent is the main OWL-OLM component which maintains the user-knowledge acquisition dialogue. The user-agent interaction is geared towards achieving the goal of the user, according to which the agent defines its dialogue goals. For example, in a learning situation the user may want to be recommended what to read next on a certain topic. The goal of the dialog agent in this case would be to asses the current state of the user's knowledge according to the requirements for the particular course task. To do so, dialog subgoals are defined (e.g. to probe the user's knowledge of concepts related to the task). In addition, the user may want to clarify some part of the domain, so he asks questions. Dialog subgoals will be then to answer the user's questions and help him clarify the particular domain aspects.

A Domain Ontology built in OWL is used to maintain the dialogue and to update the user's Short-Term Conceptual State. The latter is also represented in OWL and provides a model of the user's conceptualisation gathered throughout the dialog. The dialogue agent maintains a dialogue episode goal, which is divided into sub-goals that trigger Dialogue Games - sequences of utterances to achieve a specific sub-goal. The agent uses also a Game Analyser that analyses each user's utterance to decide the agent's response and to update the user's short-term conceptual state. When a dialogue episode finishes, the short-term conceptual state is used to update the user's Long-Term Conceptual State, referred to as the User Model.

OWL-OLM uses Jena extensively for input and output of OWL ontologies and models, creating and changing OWL resources and resolving domain ontology queries. The OWL generic reasoner from Jena is employed to make inferences from the domain ontology.
Screen shot from the OWL-OLM prototype.
The OWL-OLM screenshot shows the dialog history in the upper-left corner and the last utterance in the graphical area. The user composes utterances by constructing diagrams using basic graphical operations - create , delete or edit a concept or a relation between concepts - and selecting a sentence opener to define his intention, e.g. to answer , ask question, agree, disagree, or suggest topic . The interface uses the open source Java graph component JGraph to present, create, and modify utterances in a graphical way.

Publications

Denaux, R., Aroyo, L., & Dimitrova, V. (2005). An Approach for Ontology-based Elicitation of User Models to Enable Personalization on the Semantic Web, 14th International World Wide Web Conference , Chiba, Japan, May 2005 (poster).

Denaux, R., Aroyo, L., & Dimitrova, V. (2005). OWL-OLM: Interactive Ontology-based Elicitation of User Models, Workshop on Personalisation for the Semantic Web PerSWeb05 at 10th International Conference on User Modeling UM2005 , Edinburgh, UK, 23-29 July 2005.

Denaux, R., Dimitrova, V. , & Aroyo, L. (2005). Integrating Open User Modeling and Learning Content Management for the Semantic Web, 10th International Conference on User Modeling UM2005 , Edinburgh, UK, 23-29 July 2005.

Denaux, R., Dimitrova, V. , & Aroyo, L. (2004). Interactive Ontology-Based User Modeling for Personalized Learning Content Management, Workshop on Applications of Semantic Web technologies for E-Learning (SW-EL) held in conjunction with Adaptive Hypermedia 2004 .


Last Updated by Vania Dimitrova on 1st November 2005.