University of Surrey researcher on explainable AI for speaker recognition and speaker-embedding interpretation
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
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—
No output score yet. This axis measures publication volume — paper counts by venue and year — but the underlying bibliometric record is not available here; only two cataloged arXiv preprints exist, so no estimate is made.
Impact—
No impact score yet. This axis measures citation-based influence — citations, h-index, and uptake by other researchers — but no citation metrics are available in the input data; the two papers are recent preprints, so influence cannot yet be assessed.
Focus0/100
The focus score of 0 reflects an evenly split field distribution across the two cataloged papers: 50% 'speech processing / speaker recognition (explainable AI)' and 50% 'speech technology and machine learning (explainable AI)'. In practice the two labelled fields are closely allied, both centred on explainable AI for speaker recognition, so the zero concentration score says more about the even 50/50 split and the very small two-paper sample than about genuine topic dispersion.
Collaboration22/100
Collaboration scores 22 out of 100, based on only 2 distinct co-authors in the cataloged record. The recurring partners are Wenwu Wang and Mark D. Plumbley, and Yanze Xu takes first authorship on both papers (corresponding-author flag not set on either). The pattern suggests a tight, mentorship-style team at the University of Surrey rather than a broad network; the small two-paper sample means this measure may change quickly as more publications are indexed.
Activity—
No activity score yet. This axis measures publishing cadence over time — recent output volume and continuity — but the profile lacks a full publication history beyond the two 2026 preprints, so research tempo cannot be scored.
Reproducibility38/100
Reproducibility readiness averages 38 across the 2 papers Evolve has reviewed. The hierarchical-organisation paper is rated medium — it introduces HCCM and the L-score but evaluates a single model and dataset without clustering baselines — while the second-order-pattern paper is rated low, with the Gradient Cost criterion described as heuristic and the evaluation thin. A score in the high-30s therefore signals methods that are documented but not yet independently re-runnable without extra experimental detail.
Research summary
Yanze Xu is a researcher at the University of Surrey working at the intersection of speaker recognition and explainable AI. The Evolve catalog contains two first-author arXiv preprints, both dated 2026, that together outline a coherent research programme: making the internal structure of speaker embeddings interpretable through hierarchical single-linkage (SLINK) clustering. The first paper, 'Interpreting hierarchical organisation of speaker embeddings', proposes HCCM and the L-score, pairing hierarchical clustering with cluster-class matching to interpret SLINK hierarchies of speaker embeddings; Evolve's review finds the findings interesting but notes a single model/dataset and no clustering baselines. The follow-up, 'Exploring Second-Order Pattern Recognition in Speaker Recognition', extends the SLINK+HCCM framework with the new HCNA method to discover and assign latent second-order patterns; reviewers flag the Gradient Cost criterion as heuristic and the evaluation as thin. Reproducibility readiness averages 38 across the two papers (one medium, one low). Collaboration data shows a compact team: 2 distinct co-authors, with Wenwu Wang and Mark D. Plumbley as recurring partners, and Xu in first-author position on both papers. Field labels split evenly (50/50) between speech processing/speaker recognition and speech technology/machine learning, both under the explainable-AI banner. Output, impact, and activity metrics are not yet available, so this should be read as an early-stage profile built on a small, recent, preprint-heavy record.
Collaboration pattern
Yanze Xu's cataloged record shows a compact collaboration footprint: only 2 distinct co-authors (collaboration score 22/100), with Wenwu Wang and Mark D. Plumbley appearing as the recurring partners on both 2026 papers. Xu is first author on each, with no corresponding-author role recorded. This pattern is consistent with a close-knit, mentorship-style team at the University of Surrey; given the two-paper sample, the network picture is likely to broaden as more publications are indexed.
Research keywords
Highlights
Introducing HCCM and the L-score for interpreting SLINK hierarchies of speaker embeddings
Discovering latent second-order patterns in speaker embeddings with HCNA
A two-paper methodological line on explainable speaker recognition
Timeline
- 2026First cataloged first-author preprint appears on arXiv: 'Interpreting hierarchical organisation of speaker embeddings', introducing HCCM and the L-score for SLINK hierarchies of speaker embeddings.
- 2026Follow-up preprint 'Exploring Second-Order Pattern Recognition in Speaker Recognition' extends the framework with the new HCNA method to uncover latent second-order patterns.
- 2026Evolve reviews both papers, rating reproducibility readiness medium and low respectively (mean 38) and flagging single-dataset evaluation and heuristic criteria as limitations.
- 2026Both papers carry first-authorship by Yanze Xu with Wenwu Wang and Mark D. Plumbley as co-authors, defining the current 2-distinct-coauthor collaboration footprint.
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.
- The profile rests on only two cataloged papers (both 2026 arXiv preprints, first-authored), so all scores are based on a very small sample.
- Output, impact and activity dimensions lack sufficient bibliometric data and are deliberately not scored; no citation counts, h-index or publication-volume figures were available.
- Reproducibility (mean 38) reflects Evolve's review of just 2 papers; readiness ratings are qualitative (medium, low), not independent replications.
- Identity is name-resolved from public bibliographic data linked to ORCID 0000-0001-9736-5548; no biographical information beyond the listed affiliation was available.