Newer
Older
# classifying-eugeniana
1. Project phase 1 \(03-04/2021\)
>*see also description of project at [ONB Forschungsblog](https://www.onb.ac.at/forschung/forschungsblog/artikel/machine-learning-fuer-die-provenienzerschliessung).*
Primary contribution by Emanuel Zangger:
- training and testing of ML algorithm;
- set-up of pipeline for ingest of ABO barcodes and classification.
E. Zangger trained and tested on barcodes \(identifiers of digitized books\) for the publication years 1550-1599 and 1700-1738.
Additional contribution by Martin Krickl:
- initial research question
- Selection of ground truth
- evaluation of output
Extension by Simon Mayer \(ONB\) and Martin Krickl \(ONB\) for paper "Mit Machine Learning auf der Suche nach Provenienzen - ein Use Case der Bildklassifikation an der Österreichischen Nationalbibliothek" published in [Bibliothek \- Forschung und Praxis](https://www.degruyter.com/journal/key/bfup/html). \(Publication upcoming in 03/2022\).
Contribution by Simon Mayer:
- adjustments to ML algorithms trained in phase 1
- adjustment of pipeline
- testing of barcodes
- Author of paper in BFuP
Contribution by Martin Krickl:
- evaluation of output
- Author of paper in BFuP
3. This project repository contains:
- the code by E. Zangger for training the model and running the pipeline
- ABO barcodes \(and corresponding predictions of the BE model\) for the years 1550-1599 and 1700-1738.
- ABO barcodes \(and corresponding predictions of the BE models\) for the years 1501-1550 and 1600-1699.
NOTE:
*The predictions have been sorted such that the positive predictions are followed by negative predictions.*
