|
Fonctionality
Spell checking automatically corrects spelling
errors made while entering medical terms in search queries. These specialized,
infrequently used terms are the source of many common errors.
The programs underlying WRAPIN are largely
language independent, which means that only the interface need be translated
into any desired language.
Identification of the most appropriate
keywords in a document is the first, and most important step in the
production of relevant results. The main source of keywords is the MeSH
vocabulary. All other modules of the WRAPIN system depend on this all-important
function.
Two characteristics of the UMLS are utilised
in WRAPIN: the superior lexical coverage compared with MeSH, and cooccurrences,
which improves the selection of keywords.
When there are several referenced databases
to choose from it is important to choose only that/those most relevant
to the query. A selection module categorises data sources based either
on the frequency of terms occuring in the database (when the source
is available) or according to the MeSH hierarchy (in the case of remote
databases) in order to find the most appropriate resources for each
query.
Some of the remote data sources make use
of boolean logic. A module has been designed to query these sources
and extract the useful essence, to mitigate the limitations of this
type of system.
This module chooses the best documents retrieved by WRAPIN and sorts them in
order of decreasing relevance. The comparison of each document with the
query takes place at this stage.
The goal is to identify the conclusion within the text of each of the best
documents retrieved by WRAPIN. The conclusion is presented as the document
summary on the results page.
|