NLP researcher at Barcelona Supercomputing Center working on MT evaluation benchmarks and culture-aware corpus auditing
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
14
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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 scored yet: this axis would aggregate normalized publication volume and venue weight over the career record. The OpenAlex year counts (e.g. 15 works in 2025, 17 in 2026) exist, but the composite output score was not computed here.
Impact—
Not scored yet: this axis would measure citation uptake and downstream reuse of released benchmarks and corpora. Although the raw record lists h-index 14 and i10-index 23, per-paper citation linkage was insufficient to compute the score.
Focus0/100
The radar reports a focus score of 0, reflecting breadth rather than concentration: the field distribution splits evenly (0.5/0.5) between machine translation evaluation and corpus auditing/cultural evaluation resources, and OpenAlex topic counts span natural language processing techniques (116), topic modeling (73), speech recognition and synthesis (41), translation studies (39), and hand gesture recognition (17). Within NLP this reads as a deliberately wide evaluation-and-resources agenda rather than a single niche.
Collaboration83/100
Collaboration scores 83 on the strength of a wide network: 63 distinct co-authors appear in the OpenAlex record, and twelve recur often enough to be flagged — including Vilém Zouhar, Jannis Vamvas, Pinzhen Chen, Sara Papi, Ona de Gibert, and Niyati Bafna. Both 2026 catalog papers are middle-authored, multi-team efforts, and the Last Translation Benchmark is described as community-built, consistent with consortium-style benchmark construction.
Activity100/100
Activity is scored 100, the maximum. The OpenAlex year-by-year counts show an unbroken record from 2003 (1 work) through 2026 (17 works, the highest single year), with a sustained run of 10–12 works annually from 2019 to 2023, 15 works in 2025, and 38 works across the most recent three years. Output in 2025–2026 (15 and 17) clearly exceeds every earlier year in the record.
Reproducibility55/100
Reproducibility averages 55 across the two papers assessed. Both are rated medium readiness. FineWeb-CLaR releases culture, language, and region annotations for 30.9B FineWeb documents and is candid that its CT labels reach only 22% human agreement, which caps verification confidence. The Last Translation Benchmark strengthens reproducibility with 3,456 hard exceptions and rule-based verifiers, but its verdict notes reliance on an LLM verifier whose agreement is not quantified.
Research summary
Cristina España-Bonet is a researcher at the Barcelona Supercomputing Center whose profile is name-resolved from public OpenAlex bibliographic data. Her record spans 2003–2026, with 140 works summed across the years counted in the snapshot, an h-index of 14 and an i10-index of 23. Two strands dominate current work. The first is machine translation evaluation: the community-built Last Translation Benchmark (2026) assembles 3,456 hard exceptions with rule-based verifiers that separate frontier models from human rankings. The second is pretraining corpus auditing for culture and region: FineWeb-CLaR (2026) adds culture, language, and region annotations to 30.9B FineWeb documents while candidly reporting only 22% human agreement on its CT labels. The radar reflects this dual profile: activity is scored 100 (38 works in the last three years; 17 in 2026, the record's highest), collaboration 83 (63 distinct co-authors, a recurring cluster including Vilém Zouhar, Jannis Vamvas, Pinzhen Chen, and Ona de Gibert), and reproducibility 55 averaged over the two reviewed papers. Focus scores 0, consistent with an even 0.5/0.5 field split and topics ranging from topic modeling (73) to speech recognition (41) and translation studies (39). Output and impact axes are not yet scored; the paper-level evidence base is two arXiv entries from 2026, both middle-authored.
Collaboration pattern
The OpenAlex record lists 63 distinct co-authors, and twelve recur frequently enough to be flagged. The recurring set clusters around multilingual NLP and MT evaluation — Vilém Zouhar, Jannis Vamvas, Pinzhen Chen, Sara Papi, Mukund Choudhary — alongside corpus and cultural NLP colleagues such as Ona de Gibert, Niyati Bafna, and Bhavitvya Malik. Both reviewed 2026 papers are multi-author, middle-authored efforts, and one benchmark is explicitly community-built, indicating consortium-style working habits.
Research keywords
Highlights
Culture, language, and region annotations for 30.9B FineWeb documents
A community-built benchmark that separates frontier models from human rankings
A 2026 double release across both declared fields
The two reviewed papers land in the same year and map exactly onto the researcher's declared 0.5/0.5 field split: one advances machine translation evaluation, the other culture-aware corpus auditing. Both are middle-authored, non-corresponding arXiv papers rated medium on reproducibility readiness.
Timeline
- 2003First year with a recorded work in the OpenAlex counts-by-year series (1 work).
- 2019Output jumps to 12 works, the highest yearly count to that point, starting a sustained multi-year run.
- 2023Sustained pace continues with 12 works recorded, matching the 2019 peak.
- 2025Yearly count rises to 15 works, exceeding every prior year in the record.
- 2026Highest count in the record (17 works); both Evolve-reviewed papers — FineWeb-CLaR and the Last Translation Benchmark — are dated to this year.
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.
- Bibliometrics are computed from OpenAlex records only; the output and impact radar axes are not scored in this snapshot.
- The reproducibility score (55) averages only the two papers reviewed by Evolve and may not generalize to the full record.
- The source record lists the country as GB while the named institution is the Barcelona Supercomputing Center; affiliation metadata should be treated as unverified.
- Paper-level verdicts rest on two 2026 arXiv entries; no peer-reviewed venue information for them is included in the input data.