Nhan Phan

Aalto University

Unclaimed · AI-structured profileOpenAlex

Aalto speech-processing researcher building interpretable native-reference geometries for L2 pronunciation assessment

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-02) 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—

This axis is currently empty: no publication-count or productivity data is available in the input snapshot. It would measure total research output over time — number of papers per year and cumulative volume — once a fuller bibliographic record is linked.

Impact—

No citation or influence metrics are available yet, so this axis is unscored. It would measure scholarly impact — citations and uptake of the native-reference geometry methods by other groups — and cannot be estimated from the two catalogued 2026 preprints alone.

Focus0/100

The radar focus score is 0, and the raw field distribution is a 0.5/0.5 split between two labels — speech processing/spoken language assessment (computational phonetics) and speech processing/computational linguistics (computer-assisted language learning). Both describe the same working topic, L2 pronunciation assessment, so the low score reflects label fragmentation across adjacent subfields rather than genuinely scattered research interests.

Collaboration39/100

With a collaboration score of 39, this profile shows a compact working circle: 6 distinct co-authors recorded, with Tina Raissi, Mikko Kurimo and Chenxiao Wang recurring across the catalogued work. On both 2026 papers Nhan Phan sits in a middle authorship position and is not the corresponding author, consistent with contributing inside a small, stable Aalto-based speech-processing group rather than leading large distributed teams.

Activity—

Activity data is insufficient in the current snapshot, so no score is given. This axis would measure ongoing publishing cadence and recency — whether new preprints continue appearing after the two 2026 entries — and requires a longer observation window.

Reproducibility78/100

Reproducibility is the strongest recorded axis, scoring 78 as the mean over the 2 reviewed papers (reproMean 78, reproPapers 2). The 2026 phone-class geometry paper carries a high repro-readiness flag, while the earlier coordinate-geometry paper is rated medium, partly because tuning was done on train/dev only with no held-out test set. Both papers centre on interpretable, training-free methods built on self-supervised speech representations.

Research summary

Nhan Phan is a speech-processing researcher at Aalto University whose catalogued work centres on second-language (L2) pronunciation assessment using self-supervised speech models. Two 2026 arXiv preprints define an identifiable research line: a native-reference coordinate geometry that tracks L2 English proficiency (ρ≈−0.5) without training or fine-tuning, and a follow-up phone-class geometry whose distances correlate negatively with proficiency (ρ −0.53/−0.34). Evolve's review characterises both as sound and honestly caveated, while noting open questions — much of the signal in the phone-class work overlaps basic fluency features, and the coordinate geometry was tuned on train/dev only with no held-out test. The bibliometric radar records a collaboration score of 39 built on 6 distinct co-authors, with Tina Raissi, Mikko Kurimo and Chenxiao Wang recurring; on both papers the researcher holds a middle, non-corresponding authorship position. Reproducibility is the strongest recorded dimension at 78 (mean over the two reviewed papers, one rated high and one medium for repro-readiness). The focus axis scores 0 from a 50/50 split between two adjacent field labels — computational phonetics and computer-assisted language learning — which in practice describe the same working topic. Output, impact and activity axes are unscored because the underlying bibliometric data is not yet available; this is therefore an early-stage profile that will firm up as citation and productivity records accumulate.

Collaboration pattern

Collaboration is concentrated in a small Aalto-based circle: 6 distinct co-authors underlie the radar's collaboration score of 39, with Tina Raissi, Mikko Kurimo and Chenxiao Wang recurring across both 2026 preprints. Nhan Phan appears in middle authorship positions and is not the corresponding author on either paper, consistent with contributing specialist expertise within a stable speech-processing group rather than leading multi-institution consortia.

Research keywords

second-language pronunciation assessmentautomatic speaking assessmentself-supervised speech representationsWavLMphone-class centroidssingular value decompositionGoodness of PronunciationCEFR speaking proficiencycomputational phoneticscomputer-assisted language learningphone-class embeddingstraining-free interpretable methods

Highlights

Training-free native-reference geometry tracks L2 proficiency
The 2026 coordinate-geometry paper builds an SVD subspace geometry over phone-class embeddings from self-supervised speech models, correlating with L2 English proficiency at ρ≈−0.5 without any training or fine-tuning; Evolve rates it sound and honest, with medium repro-readiness.
Phone-class geometry with high reproducibility readiness
A follow-up preprint defines interpretable distances between phone-class centroids, which correlate negatively with L2 proficiency (ρ −0.53/−0.34) while also overlapping basic fluency features; it carries Evolve's high repro-readiness flag, feeding the 78 reproducibility score.
Honest limitations shape the research line
Both preprints are flagged for evaluation caveats — tuning on train/dev only with no held-out test, and signal overlap with basic fluency features — an openly acknowledged boundary that positions the native-reference geometry work for stronger validation in future iterations.

Timeline

2026First catalogued preprint: a native-reference coordinate geometry using self-supervised speech models tracks L2 English proficiency (ρ≈−0.5), rated medium for repro-readiness.
2026Follow-up preprint refines the idea into a phone-class geometry, with distances correlating negatively with proficiency (ρ −0.53/−0.34) and a high repro-readiness rating.
2026Both papers appear within a single year with middle, non-corresponding authorship alongside Tina Raissi, Mikko Kurimo and Chenxiao Wang, consolidating a training-free, interpretable line on automatic speaking assessment.

Papers in the Evolve catalog

Frequent co-authors

Tina Raissi×2 Mikko Kurimo×2 Chenxiao Wang×1 Tina Raissi×1 Nhan Phan×1 Mikko Kurimo×1

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 39, focus 0, reproducibility 78) are computed from public bibliometric data; no citation or venue metrics were available.
Output, impact and activity axes are unscored because the underlying data was insufficient in this snapshot.
The profile rests on only two 2026 arXiv preprints; no peer-reviewed venue, ORCID or country data was provided.
'Nhan Phan' is a relatively common Vietnamese name; this profile is name-resolved from public bibliographic data and identity should be confirmed via ORCID.