Survey of Knowledge Representation and Reasoning Systems.
Scientific Publication
- Report Number:
- DSTO-TR-2324
- Authors:
- Trentelman, K.
- Issue Date:
- 2009-07
- AR Number:
- AR-014-588
- Classification:
- Unclassified
- Report Type:
- Technical Report
- Division:
- Command, Control, Communication and Intelligence Division (C3ID)
- Release Authority:
- Chief, Command, Control, Communication and Intelligence Division
- Task Sponsor:
- Exec Dir CTSTC
- Task Number:
- NS07/201
- File Number:
- 2009/1059398/1
- Pages:
- 49
- References:
- 65
- Terms:
- Artificial intelligence; Expert Systems; Knowledge representation; Logic; Reasoning; Semantic web
- URI:
- http://hdl.handle.net/1947/9996
Abstract
As part of the information fusion task we wish to automatically fuse information derived from the text extraction process with data from a structured knowledge base. This process will involve resolving, aggregating, integrating and abstracting information - via the methodologies of Knowledge Representation and Reasoning - into a single comprehensive description of an individual or event. This report surveys the key principles underlying research in the field of Knowledge Representation and Reasoning. It represents an initial step in deciding upon a Knowledge Representation and Reasoning system for our information fusion task.
Executive Summary
As part of the information fusion task we wish to automatically fuse information derived from the text extraction process with data from a structured knowledge base. This process will involve resolving, aggregating, integrating and abstracting information - via the methodologies of Knowledge Representation and Reasoning - into a single comprehensive description of an individual or event. This report surveys the key principles underlying research in the field of Knowledge Representation and Reasoning. It represents an initial step in deciding upon a Knowledge Representation and Reasoning system for our information fusion task. We find that although first-order logic is a highly expressive knowledge representation language, a major drawback of the logic as a Knowledge Representation and Reasoning system for our information fusion task is its undecidability. Moreover, most first-order automated theorem provers are not designed for large knowledge-based applications. Modal logics are gradually receiving more attention by the Artificial Intelligence community, but research in modal logics for knowledge representation still has a long way to go. A production rule (expert) system is viable as a Knowledge Representation and Reasoning system, but these systems are optimally suited for small, specific domains. To build an intelligence expert system we would require expert knowledge in pretty much everything. Frame systems are limited in their expressiveness, and moreover - in regards to knowledge representation - have been superceded by description logics. Semantic networks are excellent for taxonomies, but are not particularly suitable for our information fusion task. On a more positive note, description logics are currently very popular and are actively being researched. They are (in the most part) decidable and their open-world semantics would allow us to represent incomplete information. A further advantage is the availability of Semantic Web technologies. Description logics are still limited however; for our task, we’d need to look at very expressive logics which might lose us decidability.