ML Anthology

Generating Error Candidates for Assigning Blame in a Knowledge Base

Michael A. Weintraub, Tom Bylander
ICML 1991 pp. 33-37
doi:10.1016/B978-1-55860-200-7.50011-8 /icml/1991/weintraub1991icml-generating/

Abstract

The problem of identifying why a knowledge-based system (KBS) made a mistake and correcting it is the knowledge base refinement problem. This paper concentrates on the problem of credit assignment, determining which elements of a KBS are responsible for a mistake. One method for solving this problem is to propose a set of error candidates that would account for the KBS' error and then select the best set of error candidates. In this paper, we describe how a task description of a KBS is used to generate possible error candidates that would explain how the mistake occurred. We illustrate this point using generic tasks, and show how generic tasks can be represented using device models.

ICML Semantic Scholar

Cite

Text

Weintraub and Bylander. "Generating Error Candidates for Assigning Blame in a Knowledge Base." International Conference on Machine Learning, 1991. doi:10.1016/B978-1-55860-200-7.50011-8

Markdown

[Weintraub and Bylander. "Generating Error Candidates for Assigning Blame in a Knowledge Base." International Conference on Machine Learning, 1991.](https://mlanthology.org/icml/1991/weintraub1991icml-generating/) doi:10.1016/B978-1-55860-200-7.50011-8

BibTeX

@inproceedings{weintraub1991icml-generating,
  title     = {{Generating Error Candidates for Assigning Blame in a Knowledge Base}},
  author    = {Weintraub, Michael A. and Bylander, Tom},
  booktitle = {International Conference on Machine Learning},
  year      = {1991},
  pages     = {33-37},
  doi       = {10.1016/B978-1-55860-200-7.50011-8},
  url       = {https://mlanthology.org/icml/1991/weintraub1991icml-generating/}
}