Maastricht University NLP researcher focused on SpeechLLM alignment and attention-head interpretability
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
4
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-28) 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—
Output scoring is not yet available in this snapshot; the axis would aggregate total publication volume, venue distribution, and full-text availability to quantify how much reviewed research this profile has produced.
Impact—
The citation record is too thin to score impact reliably; this axis would track total citations, citing works, and uptake of the researcher's methods by follow-up studies across the literature.
Focus100/100
A perfect focus score reflects a profile concentrated in a single indexed field, speech and natural language processing, with topic modeling and NLP techniques tied at 12 papers apiece as the dominant clusters; a health-and-ageing strand adds dementia and cognitive impairment research (7), technology use by older adults (5), and telemedicine implementation (4) without diluting the core.
Collaboration39/100
A collaboration score of 39 rests on six distinct co-authors in the record, with Gerasimos Spanakis the most frequent partner and Jan Scholtes, Abderrahmane Issam, and Paweł Mąka recurring; both catalogued 2026 papers place the researcher in a middle author position without corresponding role, consistent with a small, stable core team rather than large consortium work.
Activity93/100
Activity scores 93 on the strength of 23 works in the last three years; the yearly record climbs from single works in 2020, 2021, and 2023 and three in 2022 to nine in 2024 and seven in each of 2025 and 2026, marking a sustained high-tempo phase concentrated in 2024–2026.
Reproducibility100/100
A perfect reproducibility mean of 100 is computed from one reviewed paper, so the ceiling score rests on a thin base; supporting signals point the same way — the zero-shot SpeechLLM alignment paper is rated high readiness while the attention-head attribution study is medium, and Evolve notes the SpeechLLM headline claims rest on single-seed runs.
Research summary
Yusuf Can Semerci is a natural language processing researcher at Maastricht University (NL) whose work centres on making large language and speech models more efficient and more interpretable. The bibliometric record, resolved from OpenAlex, shows a steep productivity ramp: yearly output rises from single works in 2020, 2021, and 2023 and three in 2022 to nine in 2024 and seven in each of 2025 and 2026, for 23 works over the last three years — the basis of an activity score of 93. Topic distribution is tightly concentrated (focus score 100) around topic modelling and NLP techniques (12 papers each), with a health-and-ageing strand covering dementia and cognitive impairment research (7), technology use by older adults (5), and telemedicine implementation (4). Basic impact indicators remain early-career in scale (h-index 4, i10-index 3), and Evolve's output and impact dimensions are not yet scored. Two 2026 arXiv papers carry full Evolve reviews: a gradient-based attribution method that locates causal attention-head relations in context-aware machine translation at over 96% less compute, and a DTW-aligned CIF plus distillation recipe for cheap zero-shot SpeechLLMs. Collaboration is compact (score 39, six distinct co-authors), anchored by Gerasimos Spanakis, with Jan Scholtes, Abderrahmane Issam, and Paweł Mąka recurring.
Collaboration pattern
The record shows a compact collaboration footprint: six distinct co-authors, anchored by Gerasimos Spanakis as the most frequent partner, with Jan Scholtes, Abderrahmane Issam, and Paweł Mąka recurring. Both 2026 catalog papers place Semerci in a middle author position without corresponding role, suggesting a contributing specialist within a small Maastricht-centered NLP group rather than a large multi-institution consortium.
Research keywords
Highlights
Causal attention-head analysis at over 96% less compute
Efficient zero-shot SpeechLLMs via token-level alignment
Timeline
- 2020First year with a recorded publication in the OpenAlex-derived record (1 work).
- 2022Three works recorded — the highest yearly count of the early 2020–2023 phase.
- 2024Output jumps to nine works, the start of a sustained high-activity phase.
- 2025Seven works recorded, maintaining the high-tempo publication pace.
- 2026Seven works so far, including two arXiv preprints fully reviewed in the Evolve catalog.
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 scores derive from OpenAlex identity resolution; output and impact dimensions are not yet scored for this profile.
- Only two papers carry full Evolve reviews; the remaining recent works appear only in aggregate counts (23 works in the last three years).
- The reproducibility score of 100 is based on a single reviewed paper (reproPapers: 1) and may not generalize across the full record.
- Yearly counts and topic frequencies come from the provided snapshot and may miss venues outside its coverage.