How can millions of personal names help us understand the linguistic and cultural diversity of an entire country? A new open-access dataset from the Democratic Republic of the Congo offers researchers an important resource for exploring this question.
Published on 8 July 2026 in the journal Data, the study “CongoNames Corpus: A Large-Scale Labeled Dataset of Congolese Personal Names” introduces a collection of more than 8 million personal-name records, compiled from national secondary-school examination results.
The corpus was developed by Tshabu Ngandu Bernard, Cansa Kayembe Amaury and Mpyana Mwamba Merlec, researchers affiliated with Université Nouveaux Horizons in Lubumbashi and Korea University in Seoul.
The authors present CongoNames as the first large-scale, openly available structured corpus of personal names from the Democratic Republic of the Congo (DRC). Its significance extends beyond computational linguistics: it provides new opportunities for quantitative anthroponymy, sociolinguistics, language-contact research and the geographical study of personal names.
A Database Covering 16 Years
CongoNames is based on the Examen d'État, the national secondary-school examination in the DRC. The researchers processed official examination result lists (palmarès) published by the country's Ministry of Education.
The resulting corpus contains:
| Feature | Coverage |
|---|---|
| Personal-name records | 8,053,983 |
| Distinct complete name strings (normalised) | 6,370,887 |
| Examination years | 2008–2023 |
| Provincial categories | 12 |
| Sub-provincial regions | 304 |
| Original PDF documents | 469 |
Each record includes the candidate's complete personal name, an administrative sex marker (M/F), an examination year and regional information.
An important distinction is that the corpus contains more than eight million records, not eight million different names or surnames. Its uniqueness statistics concern complete personal-name strings, rather than separately identified given names or family names.
Approximately 78.5% of the records contain three name tokens, while the remainder contain two, four or more tokens. This distribution offers a useful starting point for investigating the structure of Congolese personal names.
What Can the Names Tell Us?
The Democratic Republic of the Congo is among the world's most linguistically diverse countries. Its naming traditions reflect interactions between indigenous African languages, particularly Bantu languages, French and Christian religious traditions.
The authors identify the coexistence of locally rooted Bantu elements and Christian or French-origin components as an important characteristic of contemporary Congolese personal naming.
From an onomastic perspective, the availability of such a large corpus creates opportunities to investigate several questions.
How do name structures vary between regions? Which name components occur across different linguistic areas? How do indigenous naming patterns coexist with names introduced through Christianity or colonial administration? And to what extent do successive examination cohorts reveal changes in the popularity and distribution of personal-name forms?
The study already provides descriptive analyses of name lengths, token distributions, character sequences, regional diversity and similarities between provincial name inventories.
Particularly interesting is the comparison of provincial name inventories using Jaccard similarity, a measure of the overlap between two sets. The authors report generally limited overlap between complete-name inventories from different provinces.
This result is intriguing, although complete-name comparisons must be interpreted cautiously: two regions may share many common given names or surname components while having relatively few identical full-name combinations.
From PDF Documents to Searchable Data
One notable feature of CongoNames is its transparent computational methodology.
Rather than relying on manually annotated lists, the researchers developed a reproducible processing system that extracts personal names from official PDF documents and converts them into structured CSV files.
The process follows three stages:
Bronze: collection of the original examination documents and extraction of their text.
Silver: cleaning and standardisation of the extracted material, including the handling of document formatting and inconsistent spacing.
Gold: production of structured datasets for statistical and linguistic analysis.
The authors validated their extraction procedure against examination statistics contained in the original documents. Reported extraction errors were below 2% for most years, although the figure reached 7.81% in 2023, when the original document layout changed.
The dataset and processing code are openly accessible, allowing other researchers to inspect the methodology and reproduce or extend the analyses.
Why CongoNames Matters for Onomastics
Large-scale name databases have become increasingly important for the study of anthroponymic variation. Yet the geographical distribution of available resources remains highly uneven.
European and North American naming systems have been the subject of extensive historical, statistical and computational investigation. Comparable structured resources for many African countries remain much less developed.
CongoNames therefore represents an important contribution to the expansion of empirical African anthroponymy.
Its potential applications include the study of personal-name morphology, regional naming traditions, lexical and phonotactic patterns, variation across examination cohorts and the relationship between naming practices and language contact.
The corpus could also help researchers develop better tools for recognising African personal names in digital texts, an area in which widely used natural language processing systems often have limited coverage.
For historical and comparative onomastics, the next challenge will be to connect quantitative patterns with linguistic interpretation.
A recurrent name component is not necessarily a surname, and the geographical concentration of a form does not establish its etymology. The identification of individual name functions, linguistic origins and historical motivations requires additional analysis.
The dataset's three-token classification, for example, is based on the number of words in a complete name. It should not be interpreted as a verified segmentation into given name, middle name and surname.
Important Limitations
The authors acknowledge several limitations that researchers should consider.
First, the corpus represents secondary-school examination candidates, not the entire population of the Democratic Republic of the Congo. Its geographical and chronological distributions may therefore reflect differences in educational access, examination participation and source availability.
Second, regional information does not establish ethnicity or the language spoken by an individual. The database contains no verified ethnic or linguistic-origin labels.
Third, the administrative sex marker should not be confused with a person's gender identity, and the authors discourage using personal names for individual-level sex or gender prediction.
Finally, the records originate from examination results involving many young people. Although the original documents were publicly released, the researchers emphasise that large-scale access creates ethical responsibilities. Aggregated linguistic research is preferable to attempts to identify individuals or link records with external personal information.
These qualifications are particularly important for onomastic research, where the interpretation of names can otherwise lead to unsupported assumptions about identity, ancestry or cultural affiliation.
An Important Step Forward for African Name Research
The publication of CongoNames marks a valuable development in the availability of African anthroponymic data.
Its principal achievement is the creation of a large, documented and reproducible corpus that can support future linguistic investigations. The database does not, by itself, explain the origins of Congolese personal names. It does, however, provide an empirical foundation on which more detailed morphological, etymological, sociolinguistic and geographical analyses can be built.
For the international onomastic community, CongoNames also highlights the benefits of cooperation between computer scientists, linguists, historians and specialists in African languages.
As more historical records and administrative documents become available in structured form, comparable projects could gradually transform our understanding of personal naming across the African continent.
CongoNames is a welcome addition to the growing infrastructure of digital onomastics - and a reminder that the study of names has much to gain from regions whose linguistic diversity remains insufficiently represented in existing datasets.
Publication and Database Links
Original publication
Bernard, T. N., Amaury, C. K., & Merlec, M. M. (2026). CongoNames Corpus: A large-scale labeled dataset of Congolese personal names. Data, 11(7), Article 169.
https://doi.org/10.3390/data11070169
Open-access dataset (Zenodo):
https://doi.org/10.5281/zenodo.19809985
Source code and documentation (GitHub):
https://github.com/bernard-ng/drc-names-corpus
Keywords: African onomastics, anthroponymy, CongoNames, Democratic Republic of the Congo, personal-name databases, African languages, computational onomastics, name geography.
No comments:
Post a Comment