Date of Award

May 2014

Degree Type

Thesis

Degree Name

Master of Science

Department

Computer Science

First Advisor

Rashmi Prasad

Committee Members

Susan McRoy, Jun Zhang

Keywords

Citation Finding, Clinical and Medical Text, Information Retrieval, Pubmed Articles, Ranking Algorithms, Sentence Classification

Abstract

This thesis presents CiteFinder, a system to find relevant citations for clinicians' written content. Inclusion of citations for clinical information content makes the content more reliable through the provision of scientific articles as references, and enables clinicians to easily update their written content using new information. The proposed approach splits the content into sentences, identifies the sentences that need to be supported with citations by applying classification algorithms, and uses information retrieval and ranking techniques to extract and rank relevant citations from MEDLINE for any given sentence. Additionally, this system extracts snippets from the retrieved articles. We assessed our approach on 3,699 MEDLINE papers on the subject of "Heart Failure". We implemented multi-level and weight ranking algorithms to rank the citations. This study shows that using Journal priority and Study Design type significantly improves results obtained with the traditional approach of only using the text of articles, by approximately 63%. We also show that using the full-text, rather than just the abstract text, leads to extraction of higher quality snippets.

Download

Share

COinS