Showing papers on "Knowledge extraction published in 1978"

Toward Automatic Extraction of Cartographic Features.

Abstract: : The problem of automatically extracting map symbology from source imagery is studied. It is concluded that a great deal of geographic knowledge used by humans, who currently perform this extraction function, must be made available to machines before the function can be automated. Several geographic knowledge sources are discussed and an attempt is made to define paradigms under which knowledge can be encoded and used in the computer. An automatic cartographic feature extraction system (ACES) is sketched which represents a best framework for continuing development on this difficult problem given current achievements. A systems approach is taken with first consideration given to desired outputs and available inputs. It is concluded that input/output technology is far in advance of technology available for interpretation of the data. Emphasis is placed on the use of knowledge by ACES during automatic interpretation of imagery. Many types of knowledge typically used by humans appear difficult to engineer into automatic processes. Use of positional knowledge encoded in a geographic data base (GDB) is selected as the most promising avenue. Proposals are given for future research work in that direction. (Author)

A network representation for procedural knowledge

Abstract: Two kinds of knowledge necessary for text understanding systems that are procedural, not deciarative, are processes on the knowledge structure (e.g., path-tracing) and knowledge of concepts like "analogy" and "euphemism". The desirability of being able to manipulate procedures as data makes a case for having the same format for both declarative and procedural knowledge. This is achieved by writing procedures in a semantic net formalism. Executing such a procedure produces a network that explicitly contains the control structure as well as the data.