The Production Process
The conversion of the backfile archive of the scholarly journal literature available
through JSTOR takes place at two sites:
- JSTOR Main Office, New York, New York
- JSTOR Production, Ann Arbor, Michigan
The process begins in New York, where negotiations between JSTOR's Director
of Publisher Relations and journal publishers result in a non-exclusive license
for JSTOR to undertake the conversion of a particular title from paper to
digital format and then make it available to the scholarly community.
Physical copies of the title are first solicited directly from the publisher;
however, in the case of missing or damaged issues, JSTOR purchases copies
from a variety of periodical replacement services. JSTOR also accepts loans
or donations of
back issues needed from a variety of sources, including JSTOR participating institutions.
These copies arrive in Ann Arbor, where JSTOR production staff inventory them
to confirm that a complete run of the title exists. Through a page-by-page examination
of each issue, a publication record is created for the title. Preservation concerns
are addressed during this part of the process and scanning guidelines are created.
A production librarian, a serials specialist on the JSTOR staff, examines the
structure of each title and creates appropriate indexing guidelines, matching
the indexing specifications used by JSTOR to each individual title and article.
Then, each journal title is shipped to a contractor for scanning and data entry.
At the contractor's facility, the physical journal volumes are disbound and
separated into discrete issue units. Each page is scanned at 600 dpi resolution,
with meticulous attention to quality. Page images are checked for marks, folds,
or skewing, and are rejected if deemed unacceptable.
Subsequently, a table of contents file, which includes bibliographic citation
information and an item type identifier (full length article, book review,
advertisement, etc.) — as well as key words and abstracts if these exist
in the original publication — is keyed for each article in the journal
run. The indexing guidelines, created for each title by JSTOR production librarians,
ensure that the components of this table of contents file will facilitate
more effective searching and browsing in the digital environment. All digital
files created by the contractor — the page images and the table of contents
files — are downloaded to CD-ROM for shipment back to the JSTOR production
facility.
The final part of the JSTOR production process again happens at the JSTOR production
facilities, where files are uploaded from CD-ROM to the JSTOR file servers.
An intensive quality control process verifies both the quality of the page images
and the accuracy of the information contained in the keyed table of contents
files. Once an initial quality control check has been performed and the data
judged acceptable, each page image is then processed by optical character recognition
(OCR) software in order to create full-text for searching. Generally, the OCR
program yields a 97% accuracy rate, with some journals having OCR files that
are 99.95% accurate, or about one error in every 2,000 characters. Although this
level of accuracy is not high enough for presentation to users, it has proven
to be quite satisfactory for searching.
When further quality control reviews have been completed and the files have
been compressed using
Cartesian Perceptual
Compression, the availability of the title is announced to JSTOR participants.
Last updated February 9, 2004.
SEARCH
| TIPS
| SET PREFERENCES
| TERMS & CONDITIONS
