Lost and Found: Enriching Knowledge Graphs with “NIL” Persons from Historical Documents

Submitted by Enrico Daga on 03/11/2026 - 05:22

Tracking #: 4044-5258

Authors: 
Arianna Graciotti
Nicolas Lazzari
Enrico Daga
Valentina Presutti

Responsible editor: 
Guest Editors 2025 OD+CH

Submission type: 
Full Paper
Abstract: 
Vast community-driven knowledge graphs (KGs), such as Wikidata, are the primary reference data sources for Entity Linking (EL) applications. However, they exhibit significant coverage bias towards information that is widely popular on the Web, leading to underrepresentation of long-tail entities, particularly from non-contemporary contexts. Concurrently, the ongoing mass digitisation of cultural heritage resources reveals numerous named entities and associated knowledge that are currently missing from general-purpose KGs. Enriching such KGs with these ``NIL'' entities offers an opportunity to improve completeness and mitigate biases, such as gender disparities in the representation of historical figures. In this article, we investigate an approach based on retrieval-augmented generative AI to capture information about NIL entities and generate structured KGs suitable for integration into Wikidata. The approach is applied to the case of persons unknown to Wikidata who are mentioned in a collection of 19th-century musical periodicals. We empirically select 6 properties from Wikidata for entities of that type and create a manually annotated NIL-entities KG as the gold standard for evaluation. Through comprehensive experiments, we evaluate 6 State-of-the-Art Large Language Models (LLMs) from different vendors, combined with 6 different State-of-the-Art retrievers. Our results demonstrate significant variations in performance across model-retriever combinations, with a high accuracy for gender identification and family name, promising results for occupation and country of citizenship, and low accuracy for date of birth. We report a detailed error analysis and discuss the potential of our approach to mitigate historical bias in Wikidata.
Full PDF Version: 
PDF icon swj4044.pdf
Previous Version: 
Lost and Found: Enriching Knowledge Graphs with “NIL” Persons from Historical Documents
Tags: 
Reviewed
Long-term Stable Link to Resources: 
https://github.com/arianna-graciotti/KG_Construction_Historical_NIL_Entities

Decision/Status: 
Accept

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Célian Ringwald submitted on 23/Mar/2026
Suggestion:
Accept
Review Comment:

Overall Assessment:
I am pleased to recommend this manuscript for acceptance in the Semantic Web Journal. The authors have provided a thorough and comprehensive response to all review comments, demonstrating significant effort in improving the quality and transparency of their work.
Strengths of the Revision:
The revised manuscript successfully addresses the majority of concerns raised in the initial review:

Enhanced Clarity: The additions to Sections 3.5, 4.3.1, and figure captions significantly improve the manuscript's readability and reproducibility.
Transparent Limitations: The new Section 7 (Limitations) provides an honest and detailed discussion of methodological constraints and future research directions, which adds substantial value to the contribution.
Reproducibility: The clarifications regarding experimental choices (k=3 examples, F1 micro metrics, NIL handling) enhance the paper's reproducibility.

The work represents an original and well-executed contribution to the Semantic Web community, combining methodological innovation with rigorous experimentation and high-quality datasets.

Final Recommendation:
In my view, the paper is ready for publication. However, but if I would suggest one last minor edit: the addition of horizontal dotted reference lines at values 0.25, 0.5, and 0.75 across Figures 3 and 5. This would provide visual benchmarks that facilitate more nuanced interpretation and easier cross-comparison between them.

That said, even without this change, the manuscript meets the high standards of the Semantic Web Journal and makes a valuable contribution to the field.

Review #2
By Hang Dong submitted on 16/Apr/2026
Suggestion:
Accept
Review Comment:

The authors have addressed my comments. Thanks.

Regarding the "retrieved_context_texts", completely removing it may not be the best approach for open access. Sometimes they can be quite useful to evaluate the quality of retrieval, e.g. in RAG-based systems. Even just releasing the partial "retrieved_context_texts" as previously, that would be OK, as long as the authors can explain what approach was used to generate that column - this can be explained in the GitHub file.

Review #3
By Alessandro Mosca submitted on 04/Jun/2026
Suggestion:
Accept
Review Comment:

This is the second version of the contribution. In the first round, one of the reviewer noted that an explicit link to the QID-KG dataset was missing that has now been added in Sec. 4.2. Overall, the paper quality has improved and the authors addressed all the comments by the reviewers, which makes it ready for publication in the Special Issue.