Introduction
The Jesuit Military Chaplains project aims to highlight the apostolic and pastoral work carried out by members of the Society of Jesus on the war front in service to military personnel. The Portal to Jesuit Studies will provide access to the various databases created by scholars, in order to facilitate research on individual Jesuits and to document their activities through interactive maps.
The first project we are publishing is titled US Military Chaplains. It brings together data collected in the monographic issue of Woodstock Letters, vol. 89, no. 4 (November 1960), by Gerard F. Giblin, S.J., and is intended to expand over time through new findings that scholars will contribute as a result of their archival research.
The data included in the project cover American Jesuits who served at the front during the two World Wars and the Korean War, as recorded in the index of the above-mentioned periodical. For the database, the information was extracted mechanically via OCR, organized into the main database table with the support of AI, and then published on the portal after a formal review by a human editor. With regard to the map, AI extracted the relevant geographic references from the Jesuits’ biographical entries.
Methodology
Database
An OCR text layer was produced from the journal issue. The data were then cleaned by retaining only the pages relevant to building the database. Editorial material was excluded, with the exception of page numbers, in order to provide precise references to the original source for each Jesuit.
Next, the AI (GPT-5.2 Thinking) was asked to learn the target data model by providing a sample prompt consisting of the table populated with data for the first three Jesuits. The AI was then tasked with importing the data for the World War I pages in a csv table. A spreadsheet was exported in CSV format, cleaned of errors, and re-uploaded by pointing at mistakes in order to let the machine improve in data recognition. The same process was repeated to export and refine the data for World War II and the Korean War.
For reasons of efficiency (given the relatively small number of records in the dataset and the need to review the data before ingesting them for map creation), information found in less orderly sections of the index—such as valor awards, participation in more than one war, or casualties—was entered manually.
The resulting table includes the following fields:
- id: internal identifier for each Jesuit in this project (not visible in the front end)
- jesuitId: identifier for each Jesuit within the portal (not visible in the front end)
- Name: first and last name of the Jesuit
- Province: the Jesuit’s province of origin
- Born: date of birth
- Entered Society: acceptance date into the Society of Jesus
- Ordained: date of priestly ordination
- Bio: activities as a military chaplain in the Society of Jesus
- War: war(s) in which the Jesuit served as a military chaplain
- Casualties: injuries or casualties suffered by the Jesuit
- Casualties (info): longer description of the casualties column taken from the source
- Awards: awards received for support in military operations
- Source: source from which the information was extracted
The list of abbreviations used and the available filters are provided on the database page.
Map
The first step in creating the interactive map was to extract identifiable geographic data from each biographical entry and associate it with the corresponding Jesuit, indicating a time span whenever possible. By “identifiable geographic data,” we mean any information that the machine could recognize autonomously and reliably associate with a specific point on a map. For this reason, we did not include every movement made with the military units to which the chaplains were assigned, nor did we include their maritime itineraries.
To automate the process, we again trained the AI (GPT-5.2 Thinking) to recognize as many place names as possible with the highest attainable accuracy. Starting from the table created for the database, we produced a second table containing one row for each location mentioned in the biographical entries of the first three Jesuits. We then matched these locations to spatial coordinates via Google Maps (latitude and longitude) and added date references where available. Once extracted, the results were manually reviewed and corrected by documenting recurring errors—most often the inability to identify a specific place or to segment the text correctly. This phase required 13 prompt iterations before reaching a satisfactory output in which the extracted results (while not fully exhaustive) accurately corresponded to what appeared in the text. To support this work, the model was also instructed to cite the passage from which each piece of information had been extracted.
Due to the different ways dates were recorded in the original source, it was not possible to standardize date formats. For this reason, the date column was treated as a text field, and it was not possible to reconstruct a timeline. In order to make the mapped points easier to interpret during visualization, the ordering criterion applied to the id field followed two principles: first, records were grouped by individual Jesuit; second, within each Jesuit’s group, the information was arranged in chronological order whenever the source made this possible.
As for the automated correlation of latitude and longitude fields, the model identified (from the text) the appropriate search keys to be entered into Google Maps, from which coordinates could be derived. The resulting coordinates were verified by human review.
Out of 1,104 occurrences, the model was unable to recognize 18. These cases were generated by camps that no longer exist (e.g., Camp Grant in Rockford, Illinois; Camp Young in the California desert; or Camp Dao in the Philippines) or by sites whose names have changed over time (e.g., Camp Knight—today the Oakland Army Base in California—or Camp Harahan, also known as Camp Plauche in New Orleans).
The final table was then published on ArcGIS by Esri and used to create a dedicated project including: a hosted layer, an interactive map, and an Instant App. Additional information for each component is available on the pages of the respective projects.
ArcGIS references (by item ID):
Table: daae4ee2e7b249dc8819aef553a7b3be
Feature layer: 7b5256fae69949e48e6e9e15da7883fe
Interactive map: c2114070f85e4c3b804f6d17fb9c2eec
Instant App: 9267a0d134274aa1b0d449fabb18fc57
Project Editor
Alessandro Corsi (Institute for Advanced Jesuit Studies, Boston College)