Elzbieta Dawidek (University of Lower Silesia DSW Ideis)

DSW Ideis University

Unclaimed · AI-structured profileOpenAlex

Computational linguist at DSW Ideis University studying BPE tokenization limits in Polish inflectional morphology

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-21) 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—

No output score is available yet: this axis would measure total publication volume, venue types and output pace, but the snapshot contains only the single 2026 arXiv preprint, which is too little data to compute a meaningful score.

Impact—

No impact score is available yet: this axis would measure citations, citation velocity and scholarly uptake of the work, but no citation data was provided for the cataloged preprint, so it is deliberately left unscored.

Focus100/100

The focus dimension scores a maximal 100: every paper in the cataloged record — one 2026 arXiv preprint — sits in a single field, computational linguistics (NLP tokenization; Polish linguistics), so the field distribution assigns that field 100% of indexed output (fieldDist weight 1 of 1). With only one paper indexed, this full concentration reads as an early snapshot rather than an established specialization.

Collaboration0/100

Collaboration scores 0 because the bibliometric record lists zero distinct co-authors (distinctCoauthors: 0) and no frequent collaborators, matching the single first-authored, corresponding-author preprint. The axis measures co-authorship breadth within indexed output only; it reflects catalog coverage, and co-authored work outside this snapshot would not appear in the score.

Activity—

No activity score is available yet: this axis would measure recent publishing cadence and year-over-year consistency, but with a single 2026 record the snapshot cannot distinguish an emerging profile from a sparse one, so the axis is left unscored.

Reproducibility20/100

Reproducibility scores 20 (reproMean: 20) across one reviewed paper (reproPapers: 1). Evolve's verdict on the 2026 BPE-in-Polish preprint calls it 'conceptually rich but empirically thin and hard to replicate,' and its reproducibility readiness is rated low — so the score currently rests on a single, weakly reproducible data point rather than a track record.

Research summary

Elzbieta Dawidek is a researcher at DSW Ideis University (University of Lower Silesia) working in computational linguistics, with a cataloged focus on how byte-pair encoding (BPE) tokenization behaves in Polish, a morphologically rich, inflectional language. The Evolve catalog currently indexes one paper: a 2026 arXiv preprint, 'The Limits of BPE Tokenization in Polish: Segmentation-Flexional Forms, Grammatical Anchoring, and First-Person Stability in Inflectional Language Models,' on which Dawidek is first and corresponding author. The paper argues that BPE segmentation in Polish preserves only a frequency-driven 'statistical shadow' of grammar rather than anchored grammatical forms, connecting subword segmentation to flexional morphology and first-person stability in inflectional language models. Evolve's review finds the work conceptually rich but empirically thin and hard to replicate, and rates its reproducibility readiness low; the reproducibility dimension consequently scores 20 (reproMean 20 over one paper). The bibliometric snapshot is narrow: field distribution is 100% concentrated in a single field (focus score 100), and zero distinct co-authors are recorded (collaboration score 0), consistent with solo authorship. Output, impact and activity dimensions lack sufficient data and are unscored. This is an early-stage profile built from limited public bibliographic data; no ORCID or country metadata is available.

Collaboration pattern

The cataloged record shows fully solo work: zero distinct co-authors (distinctCoauthors: 0), an empty frequent-coauthor list, and a single first-authored, corresponding-author preprint, yielding a collaboration score of 0. This measures co-authorship within the indexed snapshot only; collaborators on non-indexed work would not appear. No recurring partners can be identified yet, and any collaboration patterns would emerge as more output is indexed.

Research keywords

BPE tokenizationPolishinflectional morphologysubword segmentationgrammatical form anchoringstatistical shadow of grammarcomputational linguisticsinflectional language modelsfirst-person stabilitysegmentation-flexional forms

Highlights

Argues BPE captures only a frequency-driven 'shadow' of Polish grammar
The 2026 first-authored, corresponding-author arXiv preprint claims that BPE tokenization in Polish preserves a frequency-driven statistical shadow of grammar rather than anchored grammatical forms — the core 'statistical shadow of grammar' thesis of the paper.
Connects subword segmentation to flexional forms and first-person stability
The preprint links segmentation-flexional forms, grammatical anchoring and first-person stability in inflectional language models, framing subword segmentation failures through the lens of Polish inflectional morphology.
Evolve review flags the work as conceptually rich but empirically thin
Evolve's verdict on the paper is 'conceptually rich but empirically thin and hard to replicate,' with reproducibility readiness rated low; this single reviewed paper sets the reproducibility mean of 20 (reproPapers: 1).

Timeline

2026Posts a first-authored, corresponding-author arXiv preprint on the limits of BPE tokenization in Polish, covering segmentation-flexional forms, grammatical anchoring and first-person stability.
2026The preprint enters the Evolve catalog and is reviewed as conceptually rich but empirically thin and hard to replicate, with low reproducibility readiness.
2026The bibliometric snapshot records a fully concentrated profile: one field at 100% of indexed output (focus 100), zero distinct co-authors (collaboration 0), and a reproducibility mean of 20 over one paper.

Papers in the Evolve catalog

2026-09-17first authorcorresponding
The Limits of BPE Tokenization in Polish: Segmentation-Flexional Forms, Grammatical Anchoring, and First-Person Stability in Inflectional Language Models

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.

Dimension scores are computed from a small bibliometric snapshot (a single cataloged paper), not a full publication record.
Output, impact and activity axes lack sufficient underlying data and are intentionally left unscored.
No ORCID or country metadata is available; the profile is name-resolved from public bibliographic data, and 'Elzbieta Dawidek' may be shared by multiple researchers.
Reproducibility rests on one reviewed paper (reproMean 20, reproPapers 1), which Evolve rated low in reproducibility readiness.