Speech-based clinical tools are increasingly deployed in multilingual settings, yet whether pathological speech markers remain geometrically separable from accent variation remains unclear. Systems may misclassify healthy non-native speakers or miss pathology in multilingual patients. We propose a four-metric clustering framework to evaluate geometric disentanglement of emotional, linguistic, and pathological speech features across six corpora and eight dataset combinations. A consistent hierarchy emerges: emotional features form the tightest clusters (Silhouette 0.250), followed by pathological (0.141) and linguistic (0.077). Confound analysis shows pathological-linguistic overlap remains below 0.21, which is above the permutation null but bounded for clinical deployment. Trustworthiness analysis confirms embedding fidelity and robustness of the geometric conclusions. Our framework provides actionable guidelines for equitable and reliable speech health systems across diverse populations.
Figures & tables
Figure 1: Topological analysis framework. Overview: multi-dimensional features (emotional R28 , linguistic R33 , pathological R16 ) undergo t-SNE embedding, followed by three branches: (A) clustering quality via four metrics, (B) bootstrap stability ( B=20 , 80% subsampling), and (C) confound detection via 2σ overlap with a permutation null in shared PCA space.
Figure 2: t-SNE embeddings ( R2 , perplexity = 30) across all eight combinations. Per-dimension Gaussian kernel density contour [ 31 ] (bandwidth=0.4, 30% maximum density isoline) highlight manifold extent; filled regions show core density. Silhouette score badges (lower right) quantify per-panel clustering quality. The emotional-dominant hierarchy is consistent across all combinations, with pathological features forming intermediate clusters and linguistic features showing the most diffuse structure.
Dimension
Silhouette ↑
DB Index ↓
CH Index ↑
Emotional ( R28 )
0.250±0.057
1.448±0.161
91±70
Pathological ( R16 )
0.141±0.012
1.859±0.212
44±32
Linguistic ( R33 )
0.077±0.016
2.665±0.364
22±17
Table 1: Per-dimension clustering quality (mean ± SD across 8 combinations). Higher Silhouette and CH values indicate better clustering; lower DB values indicate better separation. Feature overlap across sets is by design (e.g., formants in both linguistic and pathological sets) and contributes to moderate absolute Silhouette values.
Combination
Emotional
Linguistic
Pathological
RAV–L2A–UAS
0.964
0.814
0.926
RAV–L2A–MDV
0.888
0.799
0.826
RAV–GMU–UAS
0.964
0.811
0.926
RAV–GMU–MDV
0.888
0.812
0.826
IEM–L2A–UAS
0.949
0.814
0.926
IEM–L2A–MDV
0.848
0.799
0.826
Table 2: t-SNE trustworthiness ( k=15 ) per dimension and combination. Values indicate the proportion of high-dimensional neighbours preserved in 2D. All values exceed 0.79. The mean follows the Silhouette hierarchy: Emotional ( 0.912 ) > Pathological ( 0.876 ) > Linguistic ( 0.809 ).
Figure 3: Pathological–linguistic overlap (Eq. 6) across eight combinations, versus permutation null. The shaded region marks the 90% confidence interval of the permutation null [ 32 ] ( μnull≈0.06 , dotted). All observed values exceed the null baseline, confirming genuine shared structure, yet remain bounded ( <0.21 ). The solid line denotes the observed mean ( μobs=0.169 ). L2A pairings show higher overlap than GMU pairings, suggesting that accent diversity improves separation.
Representation disentanglement studies how to separate distinct factors of variation in learned features so models can interpret, control, and transfer them.
University of Maryland, College Park, USA · Sylhet MAG Osmani Medical College Hospital, Bangladesh · Dr. M R Khan Shishu Hospital & Institute of Child Health, Bangladesh +1