Withdraw
Loading…
An Information-Theoretic Model as Predictor of Medical Subject Heading (MeSH) Revisions
Jackson, Larry S.
Loading…
Permalink
https://hdl.handle.net/2142/34616
Description
- Title
- An Information-Theoretic Model as Predictor of Medical Subject Heading (MeSH) Revisions
- Author(s)
- Jackson, Larry S.
- Issue Date
- 2002-12-26
- Keyword(s)
- medical subject heading
- MeSH
- classification taxonomy
- terminology evolution
- emerging concepts
- Abstract
- "Maintenance of a classificatory taxonomy is a continued, necessary expense, in light of continual changes occurring in the literature of the collection. In medical literature, the US National Library of Medicine continually expends a great deal of effort in the maintenance of the MEDLINE® database and associated Medical Subject Headings (MeSH)®. There may be statistical clues or features in the literature within the collection that could be used to inform managers of upcoming needs for MeSH re-evaluation in regions where medical and scientific understandings are changing. Such an alerting mechanism may allow reviews to begin earlier on the timeline of change, before the volume of material classified according to a taxonomy that is becoming outdated grows unfortunately large. More appropriate headings may be able to be deployed earlier. A small number of MeSH entries that have undergone interesting transitions were examined by this study. MEDLINE literature was examined for a period spanning six years either side of the MeSH revisions. This investigation produced a somewhat qualitative overview of the nature of the localized changes in MeSH through the use of information theoretic measures. These measures characterize the selection of an evolving MeSH entry by characterizing the selection between this entry, its parent, and its siblings, as conceptually equivalent receipt of a particular ""message"" from a classification machine. That is, given that a classificatory message will be issued, based upon the immediate subtree of interest concerning the arrival of a new MEDLINE record, what are the probabilities associated with each of the choices? The entropy of such a message was then estimated, given the probabilities of occurrence observed to prevail over time. Both the immediate subtree, and the subtrees rooted at one level superordinate, produced plots showing unusual changes in entropy just preceding the administratively motivated reclassification events in question. Change in entropy of the annual assignment of MeSH entries appears to be a useful predictor of a need for managerial review of the associated region of the classification tree(s). The ability to discern entropy shifts in the superordinate subtree appears valuable in cases where the immediate subtree in question has very few sibling choices. The method can simply be run on the next larger definable region, should the immediate region be trivially defined."
- Series/Report Name or Number
- ISRN UIUCLIS--2002/8+UPK
- Type of Resource
- text
- Language
- en
- Permalink
- http://hdl.handle.net/2142/34616
Owning Collections
Faculty and Staff Research and Scholarship - Information Sciences PRIMARY
Articles, papers, and other research and scholarship from iSchool faculty and staffManage Files
Loading…
Edit Collection Membership
Loading…
Edit Metadata
Loading…
Edit Properties
Loading…
Embargoes
Loading…