Machine Learning in the Combination of Expert Opinion Approach to IR
Abstract
This paper describes a machine learning version of Robertson, Maron, and Cooper's (RMC) unified model of probabilistic information retrieval (PIR), Model 3 (1982). This version, based on the statistical technique of the combination of expert opinion (CEO), incorporates learning from human indexers and searchers. Using relevance feedback from searchers, indexers* estimates are calibrated and updated. The relevance feedback is also used by an evaluation function to weight the contributions of the different retrieval models to the combined probability of relevance used to rank retrieved documents both within the context of a single search of several iterations and across searches.
Cite
Text
Thompson. "Machine Learning in the Combination of Expert Opinion Approach to IR." International Conference on Machine Learning, 1991. doi:10.1016/B978-1-55860-200-7.50057-XMarkdown
[Thompson. "Machine Learning in the Combination of Expert Opinion Approach to IR." International Conference on Machine Learning, 1991.](https://mlanthology.org/icml/1991/thompson1991icml-machine/) doi:10.1016/B978-1-55860-200-7.50057-XBibTeX
@inproceedings{thompson1991icml-machine,
title = {{Machine Learning in the Combination of Expert Opinion Approach to IR}},
author = {Thompson, Paul},
booktitle = {International Conference on Machine Learning},
year = {1991},
pages = {270-274},
doi = {10.1016/B978-1-55860-200-7.50057-X},
url = {https://mlanthology.org/icml/1991/thompson1991icml-machine/}
}