Early-career NLP researcher at the University of Geneva building NER datasets and benchmarks for international law and global health
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: no publication-volume or output metrics were available for this record. This axis would measure total scholarly output — number of publications across venues and years — beyond the two arXiv papers Evolve has reviewed.
Impact—
Not yet scored: citation counts and influence indicators were unavailable. This axis would measure how widely the work is read, cited and reused — for example uptake of the IntLawNER benchmark — once citation data becomes available.
Focus0/100
Focus scores 0 on Evolve's 0-100 scale because the record is perfectly split: the field distribution assigns 0.5 to legal NLP and 0.5 to applied NLP for global health informatics across just two 2026 papers. With only two outputs, this even split reads as an exploratory, early-stage portfolio rather than an established niche, although both halves share the same named-entity-recognition methodology.
Collaboration32/100
Collaboration scores 32 out of 100, based on 4 distinct coauthors across the record. Both papers are first-authored with no corresponding-author role recorded, and the frequent-coauthor list — Roland Bouffanais, Didier Wernli, Antoine Geissbühler and Jean-Luc Falcone — points to a small, stable circle rather than a broad external network. For an early-stage, thesis-adjacent record, this modest footprint is typical, though coauthorship breadth has clear room to grow.
Activity—
Not yet scored: the underlying activity data — publication cadence over time — was insufficient for this record. This axis would measure research tempo, such as outputs per year and sustained engagement, once more of the publication record is captured.
Reproducibility38/100
Reproducibility averages 38/100 across the 2 papers Evolve has reviewed (reproPapers: 2, reproMean: 38). The global-health spaCy pipeline is rated low readiness, reflecting evaluations that rely on single-annotator data; IntLawNER reaches medium readiness, where a dedicated benchmark design and systematic error analysis are offset by Claude-silver annotation circularity and rare-label sparsity. The score reflects Evolve's two-paper review scope, not a full audit of released code or data.
Research summary
Genis Skura is an early-career natural language processing researcher at the University of Geneva whose 2026 record, catalogued by Evolve, consists of two first-author arXiv papers that apply the same core toolkit — named entity recognition and text classification — to two very different domains: international law and global health publications. IntLawNER contributes what Evolve's review describes as the first NER dataset and benchmark for international law, built from 2,987 sentences spanning 7 entity types, produced with an agentic, LLM-as-annotator pipeline and accompanied by sound error analysis, alongside acknowledged risks of Claude-silver circularity and rare-label sparsity. The second paper, a Geneva master's thesis, builds a spaCy NER and topic-classification pipeline over global health publications, drawing on word2vec embeddings and zero-shot and few-shot classification; Evolve calls it practical and candid while noting that its evaluation rests on single-annotator data. The field distribution is an even 50/50 split between legal NLP and applied NLP for global health informatics, the record shows 4 distinct coauthors — including Roland Bouffanais, Didier Wernli, Antoine Geissbühler and Jean-Luc Falcone — and reproducibility readiness averages 38/100 across the two reviewed papers.
Collaboration pattern
Genis Skura's catalogued work is strongly first-authored: both 2026 arXiv papers list them as first author, with no corresponding-author role recorded. The record shows 4 distinct coauthors, and the frequent-coauthor list comprises Roland Bouffanais, Didier Wernli, Antoine Geissbühler and Jean-Luc Falcone. With a collaboration score of 32, the pattern suggests early-stage work within a small, stable team rather than an extensive cross-institutional network.
Research keywords
Highlights
IntLawNER: first NER dataset and benchmark for international law
Agentic LLM-as-annotator pipeline for legal entity extraction
spaCy NER and topic classification for global health publications
Timeline
- 2026Releases IntLawNER, described by Evolve as the first named entity recognition dataset and benchmark for international law (2,987 sentences, 7 entity types), as first author on arXiv.
- 2026Publishes a first-author arXiv paper — a Geneva master's thesis — presenting a spaCy NER and topic-classification pipeline for global health publications.
- 2026The two 2026 first-author outputs split the field profile evenly between legal NLP and applied NLP for global health informatics, with 4 distinct coauthors across the record.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric dimensions are computed by Evolve from public bibliographic data; no citation, publication-count or ORCID-linked metrics were available, so output, impact and activity axes are unscored.
- Coverage is limited to the two arXiv papers reviewed by Evolve; any other publications by this researcher are outside the scope of this profile.
- The profile is name-resolved from public bibliographic data; no ORCID or country field was available, so identity disambiguation is not fully guaranteed.
- The reproducibility score (mean 38/100) reflects Evolve's review of only 2 papers, not a full audit of released code and data.