Social Sciences and Digital Humanities Archive

Metrics

66,593 Downloads

Welcome to SODHA, the Belgian federal data archive for social sciences and the digital humanities!

Here you can find and deposit social science and digital humanities data in reusable form. Published datasets receive a DOI, making them citable like other types of publications. SODHA promotes open data by enabling reuse of research data and by safely preserving datasets in the long term.

SODHA is the Belgian service provider in the Consortium of European Social Science Data Archives (CESSDA) and is hosted by the State Archives of Belgium. SODHA was built with the help of DEMO (UCLouvain) and Interface Demography (VUB).

You can consult the SODHA Guide here, and you can read our policies here.

Want to learn more about SODHA? Consult our brochure or our presentation on the State Archives' website.

If you have any question, you can contact us at sodha@arch.be.

State Archives of Belgium

Belgian Archives for Social Sciences

Royal Library of Belgium

The Social Study

NICC-INCC

Royal Museums of Fine Arts

Featured Dataverses

In order to use this feature you must have at least one published dataverse.

Publish Dataverse

Are you sure you want to publish your dataverse? Once you do so it must remain published.

Publish Dataverse

This dataverse cannot be published because the dataverse it is in has not been published.

Delete Dataverse

Are you sure you want to delete your dataverse? You cannot undelete this dataverse.

21 to 30 of 4,343 Results

02b_navez_NER_spacy_off_the_shelf_no_dk_nlp_pipe_lg.ipynb Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Jupyter Notebook - 255.7 KB - MD5: 1e35b33c8202285ec93bb0d8508ec8ff Implements Workflow 2 using the complete fr_core_news_lg pipeline. Results are compared with Workflow 1 to evaluate the effect of the full NLP pipeline on NER performance.
03_navez_NER_spacy_custom-trained_dk.ipynb Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Jupyter Notebook - 575.5 KB - MD5: 5b3ef0c547261ba0b3323ce0fd2ca50b Implements Workflow 3 described in the article. The notebook fine-tunes fr_core_news_lg on the Navez gold-standard corpus by extending the default spaCy model with the domain-specific entity types ART, EXH, GRP and LETT, in addition to the standard PER, LOC and ORG categories. It...
best_hps.json Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal JSON - 474 B - MD5: a34a656006d26bd49168cb552f29955f JSON file containing the final hyperparameter configuration selected for each transformer model (CamemBERT, CamemBERTav2, D'AlemBERT and Europeana BERT) after hyperparameter optimisation.
combined_eval_scores3.tab Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Tabular Data - 1.4 KB - 13 Variables, 24 Observations - UNF:6:IHdDRnEBZ4B81/KMQBqVKA== Consolidated spreadsheet containing the evaluation scores of all experiments performed in this study. Facilitates comparison between off-the-shelf spaCy models, the custom-trained spaCy model, and the transformer-based models across different evaluation scenarios and entity types...
cv_results.tab Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Tabular Data - 1.8 KB - 6 Variables, 20 Observations - UNF:6:72/ohGHOKcR5yXwDUXrubA== Cross-validation results for the transformer-based NER models. Reports evaluation metrics across five training runs with the best hyperparameters per model to assess model robustness.
development_set-2.csv Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Comma Separated Values - 167.4 KB - MD5: e2f0fbe2ba2975b0de463eb734e611ab Development (validation) partition of the corpus used for hyperparameter optimisation, model selection and intermediate evaluation during NER model training.
development_set-2.pkl Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Unknown - 8.2 MB - MD5: 8d104a218fd93d5b17d6bf1671104d10 Python Pickle version of the development dataset used in the training workflows.
errors_camembertav2.tab Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Tabular Data - 13.8 KB - 10 Variables, 159 Observations - UNF:6:dX1zhbyrPPk1DreziVEOHw== Prediction errors produced by the fine-tuned CamemBERTav2 model. Used for qualitative comparison with the other transformer-based NER models.
errors_camembert.tab Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Tabular Data - 43.3 KB - 10 Variables, 529 Observations - UNF:6:LKJL6zjeIoC7+p0mefydwg== Prediction errors produced by the fine-tuned CamemBERT model. Documents incorrectly recognised, partially recognised, missed and spurious entities for qualitative error analysis.
errors_dalembert.tab Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal Tabular Data - 16.0 KB - 10 Variables, 185 Observations - UNF:6:+TqzKZgt8g2h9mleixtmLw== Prediction errors produced by the fine-tuned D'AlemBERT model. Documents representative recognition errors analysed in the accompanying study.

02b_navez_NER_spacy_off_the_shelf_no_dk_nlp_pipe_lg.ipynb

Aug 25, 2026 - Replication Data for: From nineteenth-century letters to entities: "a NER pipeline for French correspondence and its methodological lessons" - Article for Digital Humanities Benelux Journal

Jupyter Notebook - 255.7 KB -

Implements Workflow 2 using the complete fr_core_news_lg pipeline. Results are compared with Workflow 1 to evaluate the effect of the full NLP pipeline on NER performance.

03_navez_NER_spacy_custom-trained_dk.ipynb

Jupyter Notebook - 575.5 KB -

Implements Workflow 3 described in the article. The notebook fine-tunes fr_core_news_lg on the Navez gold-standard corpus by extending the default spaCy model with the domain-specific entity types ART, EXH, GRP and LETT, in addition to the standard PER, LOC and ORG categories. It...

best_hps.json

JSON - 474 B -

JSON file containing the final hyperparameter configuration selected for each transformer model (CamemBERT, CamemBERTav2, D'AlemBERT and Europeana BERT) after hyperparameter optimisation.

combined_eval_scores3.tab

Tabular Data - 1.4 KB - 13 Variables, 24 Observations -

Consolidated spreadsheet containing the evaluation scores of all experiments performed in this study. Facilitates comparison between off-the-shelf spaCy models, the custom-trained spaCy model, and the transformer-based models across different evaluation scenarios and entity types...

cv_results.tab

Tabular Data - 1.8 KB - 6 Variables, 20 Observations -

Cross-validation results for the transformer-based NER models. Reports evaluation metrics across five training runs with the best hyperparameters per model to assess model robustness.

development_set-2.csv

Comma Separated Values - 167.4 KB -

Development (validation) partition of the corpus used for hyperparameter optimisation, model selection and intermediate evaluation during NER model training.

development_set-2.pkl

Unknown - 8.2 MB -

Python Pickle version of the development dataset used in the training workflows.

errors_camembertav2.tab

Tabular Data - 13.8 KB - 10 Variables, 159 Observations -

Prediction errors produced by the fine-tuned CamemBERTav2 model. Used for qualitative comparison with the other transformer-based NER models.

errors_camembert.tab

Tabular Data - 43.3 KB - 10 Variables, 529 Observations -

Prediction errors produced by the fine-tuned CamemBERT model. Documents incorrectly recognised, partially recognised, missed and spurious entities for qualitative error analysis.

errors_dalembert.tab

Tabular Data - 16.0 KB - 10 Variables, 185 Observations -

Prediction errors produced by the fine-tuned D'AlemBERT model. Documents representative recognition errors analysed in the accompanying study.

Add Data

Share Dataverse

Link Dataverse

Reset Modifications