Data Integration in an Agent- Based Framework
Dr. Maria Vargas-Vera
Project Area: Data Integration, Distributed Systems and Semantic Web
The recent popularity of the Web created a demand for software solutions that help the user to sort out relevant information from the vast number of data available in this media represented by static, dynamic web pages and Web enabled databases. Besides the popular search engines like (Yahoo, Google) a very promising solution is the question-answering (QA) system, which provides answers to specific questions raised by the user. An stepping stone is the system called AQUA, which amalgamates Natural Language Processing (NLP), Logic, Ontologies and Information retrieval techniques is envisioned to play an important role in the development towards the Semantic Web.
Considering the dynamic nature of the Semantic Web that is the extension of the current World Wide Web (WWW) it is hardly imaginable that isolated applications will be able to serve successfully the users’ ever growing requirements since the information available to human decision makers continues to grow beyond human cognitive capabilities. In such an environment, a single agent or application limited by its knowledge, perspective and its computational resources cannot cope with the before mentioned scenarios effectively. As the domain becomes larger and more complex, open, and distributed, a set of cooperating agents is needed to address the reasoning task effectively. Each agent carries only a partial knowledge representation about the domain and can observe the domain from a partial perspective where available prior knowledge is generally uncertain. Extensive study of the subtask of how multiple agents can collectively reason about the state of the domain based on their local knowledge, local observation, and limited communication needs to be carried out in order to ensure that an agent in a multi-agent system can reason and act autonomously as in the single-agent paradigm, to overcome its limit in domain knowledge, perspective, and computational resource, it can benefit from other agents’ knowledge, perspectives, and computational resources through communication and coordination.
Within the AQUA framework if a complex question is being asked by the user, a broker agent then divides the user’s question into sub queries, which are passed to specialist agents. Each agent is capable of querying a potential resource. When the specialist agents return the result back to the broker agent, the broker composes a single coherent answer to the user’s original question. In the above mentioned multi agent architecture where different agents have access to different heterogeneous, distributed sources and these agents are responsible for a domain specific area. This means that before AQUA could answer the posed query, agents need to establish mapping between their concepts and properties in their domain specific ontologies in order to provide meaningful and integrated information that corresponds to the query. In the context of the Semantic Web mapping between concepts and their relations to each other needs to be established on the fly instead of using a mediated ontology that was created beforehand. This implies that agents engage in negotiating about the concepts and their relationships that is present in their different ontologies in order to integrate the data. Building such mapping involves reasoning under uncertainty about the similarity of concepts in different ontologies. Therefore, uncertainty handling needs to be addressed in this project. In particular, for uncertainty handling we will explore a Bayesian approach.
The main research question will be to address the challenges when performing data integration. These challenges are namely contradictions and overlapings of knowledge. It is expected that the student could make possible contributions to the field of data integration given the fact that the problem of data integration has not been solved completely.
The student undertaking this project will be supervised by Dr Maria Vargas-Vera who is a researcher working in the fields of Knowledge-Based Systems, Ontology Mapping and Agent-Based System.
For further information on this project, please contact
Dr. Maria Vargas-Vera (email: email@example.com)
Copyright 2012 MVargas-Vera.org. All right reserved.