Have a personal or library account? Click to login
Machine Learning Assisted Conversion of Biographical Records Into Wikidata Triples Cover

Machine Learning Assisted Conversion of Biographical Records Into Wikidata Triples

By: Daniel Baránek  
Open Access
|Jan 2026

Figures & Tables

johd-12-466-g1.png
Figure 1

Application of the segmentation model to an unseen BLGBL page.

johd-12-466-g2.png
Figure 2

Annotated full text of the entry ‘Donath, Eduard’, BLGBL, Vol. I, Issue 4, p. 271.

Table 1

Evaluation test of the biography2wikidata model.

BASIC STATEMENTSQUALIFIER STATEMENTSTOTAL
Ground truth1,2095721,781
Valid statements produced by the model714120834
Accuracy0.59060.20980.4683
Loss0.40940.79020.5317
Table 2

Comparison of entries in the BLGBL and on Wikidata: birth and death dates.

CATEGORYBIRTH DATE (P569)DEATH DATE (P570)
Same value2,85080.2%2,75876.9%
More detailed in the BLGBL1835.1%2276.3%
Less detailed in the BLGBL862.4%651.8%
Contradictory value39811.2%45512.7%
Value only in the BLGBL381.1%802.2%
Total values3,5553,585
Table 3

Comparison of entries in the BLGBL and on Wikidata: birth and death places.

CATEGORYBIRTH PLACE (P19)DEATH PLACE (P20)
Same value2,61874.9%2,36768.2%
More detailed in the BLGBL621.8%421.2%
Less detailed in the BLGBL441.3%1644.7%
Contradictory value2025.8%1805.2%
Value only in the BLGBL56716.2%71720.7%
Total values3,4933,470
Table 4

Comparison of entries in the BLGBL matched with Wikidata (WD): non-unique properties.

PROPERTYBLGBL ENTRIESBLGBL CLAIMS
WD IDLABELWITH THE PROPERTYIN TOTALMISSING ON WD
P106occupation3,66412,9485,45442.1%
P69educated at1,9682,6261,99976.1%
P39position held1,2622,6631,48455.7%
P512academic degree1,0731,1631,09794.3%
P410military rank20966053380.8%
P551residence46251850096.5%
P937workplace1,00453446580.1%
P463membership64952645386.1%
P166award received67251543484.3%
P1066student of33433330792.2%
P3342significant person233290290100.0%
P108employer65824521989.4%
P22father52730813443.5%
P611religious order22032511435.1%
DOI: https://doi.org/10.5334/johd.466 | Journal eISSN: 2059-481X
Language: English
Submitted on: Nov 10, 2025
|
Accepted on: Dec 27, 2025
|
Published on: Jan 16, 2026
Published by: Ubiquity Press
In partnership with: Paradigm Publishing Services
Publication frequency: 1 issue per year

© 2026 Daniel Baránek, published by Ubiquity Press
This work is licensed under the Creative Commons Attribution 4.0 License.