From Data Science to Modular Workflows Changing Perspectives from Data to Platform: DBDIrl 1864-1922 Case Study
Bibliographic record
Abstract
Abstract Many historical data collections foot on handwritten documents and registers, whose consultation is often very difficult due to the conservation state of the physical artefacts, and whose comprehension is also made difficult by the handwriting, difficult to interpret, and the language used, different from the modern terminology. Therefore significant research efforts by historians, demographers, population health scientists and others have been started in the past with the aim of making such data collections digitally available, first on the basis of images and then as readily available repositories of transcribed data in electronically queryable formats. For the purpose of extracting data from the Irish Civil registers of deaths in the DBDIrl 1864-1922 project ( https://www.dbdirl.com ), an AI-ML Data Analytics Pipeline was proposed as a working approach validated on a subset of the data. However, the pipeline requires manual steps and it is not applicable as is on similar datasets without significant modifications to its inner workings. We are currently transforming this prototyped, single purpose product to a modular, fully automated workflow, intended to be used and reconfigured for new data in a low-code/no-code fashion by domain experts like historians. We explain our adopted analysis and refactoring process, illustrate it on part of the pipeline, including how we faced obstacles and handled pitfalls. We also evaluate its potential to become a methodical approach to transforming an interactive program to a fully automated process, in a low-code/no-code workflow style, that can be easily reused, reconfigured and extended to be able to tailor it to other datasets as needed.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.017 | 0.019 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.003 | 0.004 |
| Science and technology studies | 0.004 | 0.004 |
| Scholarly communication | 0.007 | 0.006 |
| Open science | 0.003 | 0.007 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.007 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".