Abstract
The success of database systems in the public and private sectors has prompted a desire to extend their facilities and expand their range so as to cover pictorial applications. The components of pictorial database models proposed in the past are reviewed and attention is focused on the choice of data model. The relational model, which has proved most popular, is now recognised as not suited to handling pictorial data. In addition, as database systems acquire more intelligence, data structure must simplify the task of inferring information which is not explicitly requested or represented. The relational model has several drawbacks from this point of view and object-oriented structures are now assuming popularity. An object-oriented approach to knowledge representation in pictorial database systems can be achieved using Conceptual Graphs in which object classes form a lattice and instances of classes can be pictorial frames. The structures of Conceptual Graphs support a technique for intelligent database inference which uses the type lattice and a set of plausible concept definitions to extract background knowledge which may be pertinent to the query. Furthermore, when information can not be found in the database they provide a mechanism for invoking special purpose functions which can manipulate the image. The strengths and weaknesses of this technique are closely examined. To follow up these ideas a Relational Picture Language is described which allows a pictorial database of conceptual graphs to be set up and subsequently interrogated. Each command of the language is explained and illustrated by examples. The implementation was written in FRIL. a new artificial intelligence language which supports uncertainty.
Original language | English |
---|---|
Qualification | Ph.D. |
Awarding Institution |
|
Publication status | Published - 1988 |
Externally published | Yes |