Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
NVL--a knowledge representation language based on semantic networks
Abstract
Taxonomic hierarchical networks or semantic networks have been widely used in representing knowledge in AI applications. Semantic networks have been the preferred form of representation in AI, rather than predicate logic because of the need to represent complex, structured knowledge. However, the formal semantics of these networks has not been dealt with adequately in the literature. In this thesis, semantic networks are described by means of a formal relational logic called NVL. The characteristic features of NVL are limitor lists and binary predicates. Limitor lists are similar to restricted quantifiers but are more expressive. Several special binary relations are used to express the key ideas of semantic networks. NVL is based on the principles of semantic networks and taxonomic reasoning. The unification and inference mechanisms of NVL have considerable inherent parallelism which makes the language suitable for parallel implementation. The current opinion in AI is that semantic networks represent a subset of first order logic. Rather than modify predicate logic by adding features of semantic networks, our approach has been to devise a new form of logic by considering the basic principles and epistemological primitives of semantic networks such as properties, class concepts, relations, and inheritance. Our method differs from many sorted logic in that there are only pre-determined types in many sorted logic whereas NVL accommodates not only type symbols but also other types formed by a conjunction of several properties. A property could be either a type symbol or involve a relation. The syntax and semantics of NVL are first presented. It is shown that special relations are required to incorporate important features of semantic networks. Rules in the knowledge base are represented by the $V$ relation which also plays an important role in deriving inferences. The (mathematical) correctness of NVL is proved and concepts of unification of lists and inference in NVL are introduced. Parallel algorithms for unification and inference are developed. NVL is based on the structured representation of knowledge and is hence more efficient than predicate logic. However, unlike other knowledge representation languages based on semantic networks, NVL has a clean, formal semantics which eliminates the confusion about what a sentence in the language really means. Finally, NVL is suitable for implementation on a highly parallel machine.
Subject Area
Computer science|Artificial intelligence
Recommended Citation
Hudli, Anand Vishwanath, "NVL--a knowledge representation language based on semantic networks" (1989). ETD collection for University of Nebraska-Lincoln. AAI8925243.
https://digitalcommons.unl.edu/dissertations/AAI8925243