5th Biennial Conference of the Canadian Society for Computational Studies of Intelligence

Using Spreading Activation to Identify Relevant Help


Online assistance programs should have the ability lo fulfill complex requests for information. We have built an assistance program for the Franz Lisp programming language in which users can enter multiple keyword queries in an unstructured form. The keywords are mapped into the semantic network database, and spreading activation is used to determine the object lo be retrieved. A many-to-many mapping between keywords and topics permits familiar words to refer to potentially unfamiliar and diverse topics; for example in Franz Lisp, the keyword 'add' is associated with CONS, APPEND and PLUS. A weighting scheme assigns a value for relatedness between keywords and objects, making 'add' most closely related to PLUS. Activation is directed by assigning weights to the topics and to the classes of links between objects.

  • 123798 bytes


Canadian Society for Computational Studies of intelligence

Downloads: 632 downloads

UMBC ebiquity