Literature Search

1
Automatically Finding and/or Updating URL's in a
Bibliography Using a KnowledgeBot Brad N.
Barker, Bill G. Felkey, Brent I. Fox Auburn
University Harrison School of Pharmacy, Auburn, AL

• Objectives
• Determine whether an application exists that can
  automate any of the following tasks
• Convert a Bibliography to a Database format
• Given a list of URLs, scan a recognized journal
  source to find distinct matches
• Compare the articles found with the original
  reference
• If it exists, evaluate them based on
• Ease of use (E)
• Quality of articles returned (Q)
• Benefit to researcher (B)
• Utility of design (U)
• If not, develop the application

• Background
• Bibliography formatting has not kept up with the
  changing technology of using the internet to
  track references. Very few bibliographies
  include the internet address (known as the URL)
  to a source for the referenced article or
  abstract.
• Manually obtaining or verifying references in an
  unpublished manuscripts bibliography over the
  internet is a manual, time-consuming task.
• This project explores the options for obtaining
  references in a more automated fashion, ideally
  at the push of a button.

Program Scope Search Database
PubMed Recommended by more than one expert and
detailed instructions were on its website on how
to generate a query and submit it to PubMed
and the other Entrez databases. (NIH, 2004)
Programming Language Visual Basic, and the
Microsoft Internet Transfer Control component,
which allows the program to input a URL, receive
the web page and save it to a hard disk. Output
A copy of all matching articles and a data file
importable into EndNote.
Literature Search Investigators performed an
extensive review of Internet sites to determine
possible applications that performed these tasks
and tested several software products from
commercial vendors (ex. Microsoft Frontpage
Tools, Copernic Agent Professional). There does
not seem to be a knowledge bot (defined as an
automated search engine) that automatically
searches knowledge databases dedicated to
reference journals, (Anon., 2004) though several
experts recommended the following databases to
search PubMed, ISIs Web of Science, Ingenta,
Current Contents, Iowa Drug Information Services,
International Pharmaceutical Abstracts and
Embase. The search uncovered that EndNote was
recognized as a standard for managing
bibliographies. In addition, EndNotes latest
version (ver. 7) allows the author to create a
custom format that includes the URL as a
hypertext Reference Link.

• Result
• No product was found that performs any of the
  tasks required, despite an extensive review of
  Internet sites to identify applicable software.
• Bibliography format standards were not standard
  enough to reliably parse an articles title.
  Input file had to be manually edited (which took
  two minutes for a 100-record file).
• The program reviewed 100 records from a manually
  edited sample bibliography.
• It found 30 of the 100 records in 94 seconds.
• 5 of the 30 did not match the references in the
  original article.
• Results of Manual Search 53 of the 100 records
  found in 6.5 hours of searching.
• Total Search Success Rate 25/53 or 47
• Once the data was imported into EndNote, custom
  styles were used to add the URL as a hypertext
  link from the title of the document for
  electronic publication, or to list the URL
  separately for paper publication.

Reference Links (URLs) in a bibliography in the
on-line version of New England Journal of
Medicines How can these be
automatically generated from a written
bibliography?
Program Description PubMeds URL search string
contains the Web sites address, the search
string, and the requested output format (XML,
EndNote, etc). Each documents abstract must be
requested by PubMeds identifier ID. Two passes
are necessary the first to send PubMed the
titles search string to get the ID and the
second to send the ID to get the abstract. The
program searches the HTML file for the original
search string. If the program cant find it, the
download is deleted and the program reports that
this particular search failed. If it is found,
the program reports success for this document,
saves the document ID and retrieves the next line
in the input file. After the program reads the
last line of the input file, it sends a final
search request to PubMed that contains all of the
document IDs. It saves the results of this
request as Endnote.txt, suitable for importing
into EndNote.

• Discussion
• The test sample size needs increasing, with more
  varied bibliographic formatting content.
• Program shows promise, especially with the
  addition of more search databases. More
  development testing is needed.
• Should be used as initial stage of manual search.

Example If the input file contained the
line Digital Library Services in Mobile
Computing. SIGMOD Record, 24(4)34-39, 1995.
The program first truncates the line at the
first period Digital Library Services in
Mobile Computing It would then replace the
spaces with s and put the line in
quotes DigitalLibraryServicesinMobileComp
uting It then creates the following search
string, which it sends to the ActiveX
component http//www.ncbi.nlm.nih.gov/entrez/que
ry.fcgi?cmdSearchdbPubMedtermDigitalLibrary
ServicesinMobileComputing Prefix for
requesting a list of documents for a particular
search string Search String

The program then stores the output of the
component in a temporary file and searches the
file for the first line that contains the
following string http//www.ncbi.nlm.nih.gov80/e
ntrez/query.fcgi?cmdRetrievedbPubMedlist_uids
The next 8 digits after this string are the
documents unique ID. The program reads the 8
digits and create another http//www.ncbi.nlm.ni
h.gov80/entrez/query.fcgi?cmdRetrievedbPubMed
list_uids12345678doptAbstractitooliconabstr s
earch URL using the unique document identifier.
Ex. For an ID of 12345678, the text search
string would be Prefix for requesting a
specific document
ID
Suffix, requesting the abstract The
program sends the new URL to the ActiveX
component and saves the returned results as an
HTML file with the original search strings name.
References Anon. (2004). BotSpots. Retrieved
2/17/2004, from http//www.botspot.com NIH.
(2004). Creating a Web Link to the Entrez
Databases. Retrieved 2/19/2004, 2004, from
http//www.ncbi.nlm.nih.gov/entrez/query/static/li
nking.html
Abstract Access Code BA101430
Abstract BA1177