The FDA Notices of Judgment digital archive was a R&D collaboration between the History of Medicine Division and the Lister Hill Center's Communications Engineering Branch to explore the potential for automated metadata creation from digitized textual collection materials. One outcome was the SPER system, software that uses machine learning tools to automatically identify and extract metadata from batches of page images, a quality control review and metadata editing function, and ingest of the metadata and page images into a digital repository.
For more information about the project and the Notices of Judgment themselves, see
Navigating the NJs
Each NJ digital object contains a metadata record generated from the case record's text, a text file of the extracted OCR for the case, and a PDF of the relevant page(s). The fields chosen are a combination basic bibliograhic-type information, such as Title, and other information the Curator felt users would find valuable such as Court Jurisdiction and Seizure Location.
Content was scanned only at the page level rather than at the individual case level as this would have been too time consuming, therefore in many instances there are additional cases in the PDF. Printing styles evolved throughout the publications' lifetime so the user will encounter several formats:
- Scenario 1: each NJ is a distinct unit whether it is one page or many pages long; we provide the full text just for that single NJ in this scenario.
- Scenario 2: multiple NJs per page; we provide the same page image for each separate case record. For workflow reasons we deemed it more expedient to NOT crop pages for each NJ, hence the user may need to scroll down a page to reach the desired NJ case.
- Scenario 3: NJ case text carries across two or more pages within the multi-NJs-per-page scenario; we provide two or more pages for a single case.
Reconstituting editions and issuances
The NJs were published in a variety of physical manifestations. The earliest cases were printed on a daily basis then bound together in sets of several thousand. Later, they were published monthly in batches ranging from roughly 50 to 100 cases in a pamphlet form.
- Scenario 1: daily issuances; there is no formal title page or other standard bibliographic conventions. Reconstituting the bound item is not possible.
- Scenario 2: title page also contains the first case text; title page is present. Users can reconstitute an issuance via the ISSUE DATES facet.
- Scenario 3: title page contains no real bibliographic data except for the publication date; title page images are absent. Users can reconstitute an issuance via the ISSUE DATES facet.
Indexes
None of the indexes were scanned. However, there are no printed cumulative indexes for the entire collection. One cumulative index exists for Food and Drug Cases 1-10,000 and another for cases 10,0001-20,000, but afterwards indexes only exist for each bound set or each monthly issuance. The metadata records provide the same access and enhanced access as do the NJ indexes. While examining the corpus structure, we realized FDA's indexers used specific data points within the NJ text. This enabled us to perform the automated metadata extraction for the PRODUCT KEYWORD and DEFENDANTS just as they appear in the printed indexes. The PRODUCT KEYWORD and DEFENDANT facets now serve as a cumulative index for the entire collection.
Evidence Files Archival Collection
The archival collection of evidence files consist of over 2,100 boxes of correspondence, legal records, lab reports, product labeling, photographs and other documentary evidence accumulated in case files by federal attorneys. They are described in a finding aid and were controlled and organized by the various Sample, S., or IS numbers rather than by the NJ case number, making it challenging to find specific cases. These control numbers are located in the Numbers metadata field and just below the NJ title in the hard copies. These numbers ARE NOT the F.D.C. numbers.
After locating the evidence control numbers, users can locate the box containing the desired case file listed in the finding aid. Many NJs have several evidence numbers and usually they will all be filed together with the first number. In many instances the case files do not exist at all -- we attempted to identify these missing files in the box list, along with instances where cases were moved to another file.
The archival collection is currently in offsite storage and any retrievals require 30 days prior notice. Contact the NLM Support Center for information about accessing the collection.
Last Reviewed: March 11, 2024