Aalto University researcher building interpretable native-reference geometry for L2 pronunciation assessment
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
32
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 metrics are not yet available for this profile: publication counts, venue types, and yearly output volume could not be computed from the current bibliometric snapshot, which covers only two reviewed preprints. This axis would measure the scale, venue quality, and regularity of research production over time.
Impact—
Impact data is unavailable: citation counts, h-index-type indicators, and influence beyond the two reviewed papers were not present in the input snapshot. This axis would measure how widely the work is cited and taken up by other researchers and by assessment practice.
Focus0/100
The focus axis scores 0, reflecting a topical profile split evenly (0.5/0.5 in the field distribution) between two adjacent areas: spoken language assessment / computational phonetics, and computational linguistics for computer-assisted language learning. Both catalogued 2026 papers target second-language pronunciation assessment, so the split reads as two methodological framings of one problem — yet the metric registers no single dominant field.
Collaboration44/100
The collaboration axis scores 39, based on 6 distinct co-authors in the snapshot. The pattern is a small, stable cluster — Tina Raissi and Nhan Phan recur across the record, with Chenxiao Wang also present — and Kurimo sits in the last author position on both catalogued 2026 papers, consistent with a senior supervisory role inside a compact research group rather than a broad external network.
Activity100/100
Activity metrics are insufficient: recent-year publication frequency and ongoing-project signals could not be derived from a snapshot that covers only two 2026 arXiv preprints. This axis would measure research tempo — output cadence, submission rhythm, and engagement — across recent years.
Reproducibility85/100
Reproducibility scores 78, the mean readiness across the 2 papers assessed (reproMean 78, reproPapers 2). The phone-class geometry paper rates high, while the coordinate-geometry paper rates medium — it is tuned on train/dev data only, with no held-out test set. The score reflects a generally transparent, re-runnable methodology with room to harden evaluation protocols.
Research summary
Mikko Kurimo is a researcher at Aalto University whose recent catalogued work sits at the intersection of speech processing, computational phonetics, and computer-assisted language learning. The two reviewed papers, both posted to arXiv in 2026, develop a native-reference geometry approach for second-language (L2) pronunciation assessment built on self-supervised speech models such as WavLM. Rather than training task-specific assessors, the methods construct phone-class centroids or coordinate subspaces from native speech — using tools such as singular value decomposition — and measure how far a learner's pronunciation deviates from that native reference. In both studies these distances correlate negatively with speaking proficiency (ρ −0.53/−0.34 for the phone-class geometry; around ρ ≈ −0.5 for the coordinate geometry), offering an interpretable alternative to trained scoring such as Goodness of Pronunciation. Evolve's reviews judge both papers sound and honestly caveated while flagging limits: much of the signal overlaps basic fluency features in one, and the other is tuned on train/dev data only with no held-out test. Kurimo appears as last author on both papers, working with a compact team that includes Tina Raissi, Nhan Phan, and Chenxiao Wang. The reproducibility axis scores 78/100 across the two assessed papers (one high, one medium readiness); output, impact, and activity metrics are not yet available, and the focus score of 0 reflects an even split between two adjacent subfields.
Collaboration pattern
The snapshot shows a compact, stable collaboration pattern: 6 distinct co-authors, with Tina Raissi and Nhan Phan recurring most often and Chenxiao Wang also appearing. Kurimo is last author on both 2026 papers — a position consistent with senior supervision — and no large multi-institution network is visible in the reviewed record. Collaboration scores 39/100, indicating a small-team working style rather than high co-authorship breadth.
Research keywords
Highlights
Training-free native-reference geometry tracks L2 English proficiency
Phone-class geometry with higher reproducibility readiness
A coherent line on interpretable pronunciation assessment
Both catalogued papers share one agenda — replacing trained Goodness-of-Pronunciation-style scoring with geometric deviation from a native reference under self-supervised speech models — with Kurimo as last author on each, alongside Tina Raissi, Nhan Phan, and Chenxiao Wang.
Timeline
- 2026Posts a native-reference coordinate geometry paper on arXiv, tracking L2 English proficiency at ρ≈−0.5 using self-supervised speech models.
- 2026Follows with a phone-class geometry paper whose distances correlate negatively with L2 proficiency (ρ −0.53/−0.34), rated high for reproducibility readiness.
- 2026Serves as last author on both catalogued papers, anchoring a native-reference geometry research line with co-authors Tina Raissi, Nhan Phan, and Chenxiao Wang.
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 (collaboration, focus, reproducibility) come from a public bibliometric snapshot; the output, impact, and activity axes have no underlying data yet.
- The Evolve catalog contains only two papers, both 2026 arXiv preprints; no peer-reviewed venue information is included in the input data.
- This profile is name-resolved from public bibliographic data (ORCID 0000-0001-5278-7974); country/region is not specified in the input.
- Correlation figures (ρ −0.53/−0.34, ρ≈−0.5) are quoted from Evolve's review verdicts, not independently verified.