Improving Explanation-Based Indexing with Empirical Learning
Abstract
Explanation-Based Indexing (EBI) is an analytical approach to determining which attributes of a case to use for indexing. The EBI process takes an inherently incomplete and inconsistent theory of causation and function, and uses it to explain which of the initial conditions of a case were relevant to the actions that were performed in that case. These relevant conditions form the primary index for retrieving the case in the future. Although the analytical approach has many advantages over purely inductive approaches to index determination, the incompleteness in the theory causes indices to be created which are overly-general (because the theory fails to explain all the relevant differences between cases). We intend to use inductive techniques, within the memory organization of index classes generated via EBI, to determine additional facts that are important in correctly retrieving a given case and then update the case index to reflect this additional information.
Cite
Text
Barletta and Kerber. "Improving Explanation-Based Indexing with Empirical Learning." International Conference on Machine Learning, 1989. doi:10.1016/B978-1-55860-036-2.50030-8Markdown
[Barletta and Kerber. "Improving Explanation-Based Indexing with Empirical Learning." International Conference on Machine Learning, 1989.](https://mlanthology.org/icml/1989/barletta1989icml-improving/) doi:10.1016/B978-1-55860-036-2.50030-8BibTeX
@inproceedings{barletta1989icml-improving,
title = {{Improving Explanation-Based Indexing with Empirical Learning}},
author = {Barletta, Ralph and Kerber, Randy},
booktitle = {International Conference on Machine Learning},
year = {1989},
pages = {84-86},
doi = {10.1016/B978-1-55860-036-2.50030-8},
url = {https://mlanthology.org/icml/1989/barletta1989icml-improving/}
}