An introduction to database systems: vol. I (4th ed.)
An introduction to database systems: vol. I (4th ed.)
Extending the database relational model to capture more meaning
ACM Transactions on Database Systems (TODS)
| Hi-index | 0.00 |
It has been discussed in a number of papers that the first order predicate logic is not an adequate theory to represent natural language semantic features. The primary reason for this inadequacy is due to the fact, that first order predicate logic does not have enough categories to make possible the definition of a one - one mapping from the set of linguistic categories to the set of logical constituents.A relational data model, referred to as Extended Semantic Model is described in (Balogh, 1986) as an alternative representation of the basic semantic features of a natural language. This data model is based on the cognitive model of semantics proposed by Jackendoff (1985) and on the RM/T system proposed by Codd (1979).In this paper the Extended Semantic Model is expanded to provide a formal representation of referentiality.A simple relationship can be specified between syntax and conceptual structure: every major phrasal constituent in the syntax of a sentence corresponds to a conceptual constituent, that belong to one of the major ontological categories, like EVENT, STATE, THING, PROPERTY, PLACE, DIRECTION, MANNER, etc. The ontological categories can be further categorized as TYPE or TOKEN. All phrases, that express TOKEN constituents are referencing “real world” concepts, phrases that express TYPE constituents are non-referential.A set of rules, called semantic well-formedness rules can be specified to capture that part of the meaning of sentences which is beyond the syntactic encoding, and also to describe relationships among conceptual constituents.In the Extended Semantic Model of a relational data model consisting of the RM/T data base, a set of operators defined on it and a set of integrity rules are chosen for the representation of semantic information. The kernel entities correspond to the ontological category TYPE-TOKENs, entity subtypes and super types describe the hierarchy in the ontological category TYPEs. The characteristic, associative and designative entities provide a way to represent relationships among the different TYPEs of ontological categories. Relational schemas and sub-schemas are used to represent ontological category TYPEs, tuples in the appropriate relations to represent TOKENs, integrity rules and operators to represent conceptual well-formedness rules.The relational representation of the semantic structure of a sentence S is defined as the mappingT: sentence → tuples of relations where T is a composite operator T = UoRoFoP. The operator P maps the sentence into its phrase structure tree. The operator F defines the following recursive mapping: F : (PC∪OC) → OC, where PC is the set of phrasal categories, OC is the set of ontological categories.The result of the mapping FoP(S), the functional structure of S specifies references to the ontological categories. The operator R generates referencing keys for each ontological construct and inserts the verbal constituents or their referencing keys into the relations representing the ontological categories. The operator U invokes the integrity rules expressing semantic well-formedness rules and updates the relations accordingly.The inverse mapping of T includes the input sentence S, together with the non-verbal semantic contents of S derived by the integrity rules representing the conceptual well-formedness rules. Thus, the mapping T is “loss-less”.The Extended Semantic Model permits the use of the so called wh-questions (what - where - how- why). Such questions are answered by performing query operations.The representation of the integrity rules in the extended semantic model also provides the means for the automatic generation of some integrity rules representing semantic well-formedness rules.