BibTeX
@MISC{Thirunarayan_semantic,
author = {Krishnaprasad Thirunarayan},
title = {Semantic},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract—Many complex information needs that arise in biomedical disciplines require exploring multiple documents in order to obtain information. While traditional information retrieval techniques that return a single ranked list of docu-ments are quite common for such tasks, they may not always be adequate. The main issue is that ranked lists typically impose a significant burden on users to filter out irrelevant documents. Additionally, users must intuitively reformulate their search query when relevant documents have not been not highly ranked. Furthermore, even after interesting documents have been selected, very few mechanisms exist that enable document-to-document transitions. In this paper, we demonstrate the utility of assertions extracted from biomedical text (called semantic predications) to facilitate retrieving relevant docu-ments for complex information needs. Our approach offers an alternative to query reformulation by establishing a framework for transitioning from one document to another. We evaluate this novel knowledge-driven approach using precision and recall metrics on the 2006 TREC Genomics Track. Keywords-semantic predications, question answering, back-ground knowledge, literature-based discovery, text mining I.
Keyphrases
back-ground knowledge enable document-to-document transition traditional information retrieval technique recall metric single ranked list semantic predication relevant docu-ments biomedical discipline biomedical text abstract many complex information novel knowledge-driven approach search query multiple document literature-based discovery significant burden main issue relevant document ranked list keywords-semantic predication trec genomics track complex information need question answering irrelevant document
