Structural aspects of medical literature retrieval

Abstract

Finding scientific literature forms a main part of every researcher's daily life. Modern tools that implement state of the art information retrieval technologies allow this retrieval process to be done from each researcher's computer without the need to physically access to the library during opening hours. Publications matching given queries are now returned within seconds but the researcher still has to wade through thousands of matching documents to find the important ones or modify the query in numerous iterative steps until they are satisfied with the results. Even more important from a humanitarian point of view is the requirement of medical practitioners to answer information needs quickly as they are trying to save lives or improve the quality of life for their patients. Missing out on key papers due to sub-optimal queries is a worry for most doctors and so the literature retrieval process is still quite repetitive and extremely time consuming for them. This thesis explores ways to improve the retrieval of medical literature by exploiting structural aspects of the documents indexed, using existing medical ontologies and structures within the immediate context of the researcher. In the case of the clinical domain this context can be represented by structured data held in the clinical records of current patients. An Intelligent Literature Retrieval Assistant relieving researchers by automating the retrieval process is described and prototypically implemented. The research prototype developed in the course of this Ph.D. studies uses properties of the patient record, such as the patient's age and recent blood test results, to derive queries, which automatically retrieve publications. The clinician can then easily alter the query to retrieve the best matching documents within a short amount of time. The design of the system aims to instantly identify the publications that are most relevant to a given patient's case rather than providing a complex interface to deeply analyse all published material. During an evaluation of the prototype, researchers expressed their appreciation for the effectiveness of the tool and some even found new publications that they were not yet aware of. -- provided by Candidate. Show more