Hugging Face's logo

Datasets:
resproj007
/
severity_comparison_results

Modalities:
Text
Formats:
json
Size:
< 1K
Libraries:
Datasets
pandas
Polars
Dataset card Data Studio Files
xet
 Community
1
resproj007 commited on
Commit
75574df
·
verified ·
1 Parent(s): 3f74eea

Add comparison README with tables

Browse files
Files changed (1) hide show
  1. README.md +27 -27
README.md CHANGED
@@ -41,8 +41,8 @@ Compares predicted severity distributions of synthetic speech methods against or
41
 
42
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
43
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
44
- | 211 | 30 | Unknown | Normal | ? | 0% | 0% | 100% | 0% |
45
- | 4014 | 30 | Unknown | Normal | ? | 3% | 0% | 97% | 0% |
46
  | F02 | 200 | Moderate | Moderate | ✓ | 31% | 68% | 0% | 0% |
47
  | F04 | 70 | Mild | Normal | ✗ | 4% | 4% | 91% | 0% |
48
  | FC02 | 100 | Normal | Normal | ✓ | 0% | 1% | 99% | 0% |
@@ -93,8 +93,8 @@ Compares predicted severity distributions of synthetic speech methods against or
93
 
94
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
95
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
96
- | 211 | 30 | Unknown | Normal | ? | 0% | 0% | 100% | 0% |
97
- | 4014 | 30 | Unknown | Normal | ? | 7% | 0% | 93% | 0% |
98
  | F02 | 182 | Moderate | Moderate | ✓ | 37% | 52% | 1% | 11% |
99
  | F04 | 69 | Mild | Normal | ✗ | 32% | 7% | 59% | 1% |
100
  | FC02 | 98 | Normal | Normal | ✓ | 5% | 12% | 83% | 0% |
@@ -106,8 +106,8 @@ Compares predicted severity distributions of synthetic speech methods against or
106
 
107
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
108
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
109
- | 211 | 30 | Unknown | Normal | ? | 0% | 0% | 100% | 0% |
110
- | 4014 | 30 | Unknown | Normal | ? | 0% | 0% | 100% | 0% |
111
  | F02 | 196 | Moderate | Moderate | ✓ | 39% | 44% | 0% | 17% |
112
  | F04 | 68 | Mild | Normal | ✗ | 18% | 3% | 78% | 1% |
113
  | FC02 | 99 | Normal | Normal | ✓ | 1% | 3% | 96% | 0% |
@@ -145,8 +145,8 @@ Compares predicted severity distributions of synthetic speech methods against or
145
 
146
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
147
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
148
- | 211 | 30 | Unknown | Normal | ? | 0% | 0% | 100% | 0% |
149
- | 4014 | 30 | Unknown | Normal | ? | 0% | 0% | 100% | 0% |
150
  | F02 | 200 | Moderate | Normal | ✗ | 0% | 2% | 84% | 14% |
151
  | F04 | 70 | Mild | Normal | ✗ | 3% | 0% | 89% | 9% |
152
  | FC02 | 100 | Normal | Normal | ✓ | 0% | 1% | 91% | 8% |
@@ -160,33 +160,33 @@ Compares predicted severity distributions of synthetic speech methods against or
160
 
161
  Compares severity distributions based on **dataset labels** (from speaker metadata) rather than classifier predictions.
162
 
163
- **Reference GT distribution**: Mild: 0.0%, Moderate: 50.0%, Normal: 32.5%, Severe: 17.5%
164
 
165
 
166
  ### GT Ranking (by Jensen-Shannon distance on labels)
167
 
168
  | Method | GT JS Dist | GT Cos Sim | GT KL Div | GT TV Dist | N | Clf Accuracy | Speaker Agreement |
169
  |--------|------------|------------|-----------|------------|---|--------------|-------------------|
170
- | original_speech | 0.0000 | 1.0000 | 0.0000 | 0.0000 | 800 | 0.6987 | 0.6667 |
171
- | spark_tts | 0.2603 | 0.8616 | 0.2364 | 0.2816 | 725 | 0.5586 | 0.2500 |
172
- | spark_knn_vc | 0.2603 | 0.8616 | 0.2364 | 0.2816 | 725 | 0.6924 | 0.8333 |
173
- | baseline_orpheus | 0.2617 | 0.8597 | 0.2389 | 0.2845 | 740 | 0.3014 | 0.3333 |
174
- | knn_vc | 0.2619 | 0.8595 | 0.2392 | 0.2851 | 739 | 0.5007 | 0.6000 |
175
- | orpheus_tts | 0.2663 | 0.8545 | 0.2487 | 0.2857 | 699 | 0.4263 | 0.6000 |
176
- | sesame_tts | 0.2663 | 0.8545 | 0.2487 | 0.2857 | 699 | 0.4621 | 0.6000 |
177
- | orpheus_knn_vc | 0.2663 | 0.8545 | 0.2487 | 0.2857 | 699 | 0.6853 | 0.5000 |
178
- | sesame_knn_vc | 0.2663 | 0.8545 | 0.2487 | 0.2857 | 699 | 0.6924 | 0.8333 |
179
 
180
  ### GT Per-Method Severity Distribution
181
 
182
  | Method | Mild | Moderate | Normal | Severe |
183
  |--------|------|----------|--------|--------|
184
- | original_speech | 0.0% | 50.0% | 32.5% | 17.5% |
185
- | spark_tts | 9.4% | 27.0% | 27.3% | 36.3% |
186
- | spark_knn_vc | 9.4% | 27.0% | 27.3% | 36.3% |
187
- | baseline_orpheus | 9.5% | 27.0% | 27.0% | 36.5% |
188
- | knn_vc | 9.5% | 27.1% | 26.9% | 36.5% |
189
- | orpheus_tts | 9.9% | 26.0% | 27.9% | 36.2% |
190
- | sesame_tts | 9.9% | 26.0% | 27.9% | 36.2% |
191
- | orpheus_knn_vc | 9.9% | 26.0% | 27.9% | 36.2% |
192
- | sesame_knn_vc | 9.9% | 26.0% | 27.9% | 36.2% |
 
41
 
42
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
43
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
44
+ | 211 | 30 | Normal | Normal | ✓ | 0% | 0% | 100% | 0% |
45
+ | 4014 | 30 | Normal | Normal | ✓ | 3% | 0% | 97% | 0% |
46
  | F02 | 200 | Moderate | Moderate | ✓ | 31% | 68% | 0% | 0% |
47
  | F04 | 70 | Mild | Normal | ✗ | 4% | 4% | 91% | 0% |
48
  | FC02 | 100 | Normal | Normal | ✓ | 0% | 1% | 99% | 0% |
 
93
 
94
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
95
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
96
+ | 211 | 30 | Normal | Normal | ✓ | 0% | 0% | 100% | 0% |
97
+ | 4014 | 30 | Normal | Normal | ✓ | 7% | 0% | 93% | 0% |
98
  | F02 | 182 | Moderate | Moderate | ✓ | 37% | 52% | 1% | 11% |
99
  | F04 | 69 | Mild | Normal | ✗ | 32% | 7% | 59% | 1% |
100
  | FC02 | 98 | Normal | Normal | ✓ | 5% | 12% | 83% | 0% |
 
106
 
107
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
108
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
109
+ | 211 | 30 | Normal | Normal | ✓ | 0% | 0% | 100% | 0% |
110
+ | 4014 | 30 | Normal | Normal | ✓ | 0% | 0% | 100% | 0% |
111
  | F02 | 196 | Moderate | Moderate | ✓ | 39% | 44% | 0% | 17% |
112
  | F04 | 68 | Mild | Normal | ✗ | 18% | 3% | 78% | 1% |
113
  | FC02 | 99 | Normal | Normal | ✓ | 1% | 3% | 96% | 0% |
 
145
 
146
  | Speaker | N | Expected | Predicted | Match | Mild | Moderate | Normal | Severe |
147
  |---------|---|----------|-----------|-------|------|----------|--------|--------|
148
+ | 211 | 30 | Normal | Normal | ✓ | 0% | 0% | 100% | 0% |
149
+ | 4014 | 30 | Normal | Normal | ✓ | 0% | 0% | 100% | 0% |
150
  | F02 | 200 | Moderate | Normal | ✗ | 0% | 2% | 84% | 14% |
151
  | F04 | 70 | Mild | Normal | ✗ | 3% | 0% | 89% | 9% |
152
  | FC02 | 100 | Normal | Normal | ✓ | 0% | 1% | 91% | 8% |
 
160
 
161
  Compares severity distributions based on **dataset labels** (from speaker metadata) rather than classifier predictions.
162
 
163
+ **Reference GT distribution**: Mild: 8.8%, Moderate: 25.0%, Normal: 32.5%, Severe: 33.8%
164
 
165
 
166
  ### GT Ranking (by Jensen-Shannon distance on labels)
167
 
168
  | Method | GT JS Dist | GT Cos Sim | GT KL Div | GT TV Dist | N | Clf Accuracy | Speaker Agreement |
169
  |--------|------------|------------|-----------|------------|---|--------------|-------------------|
170
+ | original_speech | 0.0000 | 1.0000 | 0.0000 | 0.0000 | 800 | 0.5675 | 0.7500 |
171
+ | baseline_orpheus | 0.0000 | 1.0000 | 0.0000 | 0.0000 | 800 | 0.3538 | 0.5000 |
172
+ | knn_vc | 0.0006 | 1.0000 | 0.0000 | 0.0008 | 799 | 0.5382 | 0.6000 |
173
+ | spark_tts | 0.0029 | 1.0000 | 0.0000 | 0.0037 | 785 | 0.5924 | 0.2500 |
174
+ | spark_knn_vc | 0.0029 | 1.0000 | 0.0000 | 0.0037 | 785 | 0.7159 | 0.8750 |
175
+ | orpheus_tts | 0.0110 | 0.9996 | 0.0005 | 0.0144 | 759 | 0.4717 | 0.6000 |
176
+ | sesame_tts | 0.0110 | 0.9996 | 0.0005 | 0.0144 | 759 | 0.5046 | 0.6000 |
177
+ | orpheus_knn_vc | 0.0110 | 0.9996 | 0.0005 | 0.0144 | 759 | 0.7088 | 0.5000 |
178
+ | sesame_knn_vc | 0.0110 | 0.9996 | 0.0005 | 0.0144 | 759 | 0.7141 | 0.8750 |
179
 
180
  ### GT Per-Method Severity Distribution
181
 
182
  | Method | Mild | Moderate | Normal | Severe |
183
  |--------|------|----------|--------|--------|
184
+ | original_speech | 8.8% | 25.0% | 32.5% | 33.8% |
185
+ | baseline_orpheus | 8.8% | 25.0% | 32.5% | 33.8% |
186
+ | knn_vc | 8.8% | 25.0% | 32.4% | 33.8% |
187
+ | spark_tts | 8.7% | 25.0% | 32.9% | 33.5% |
188
+ | spark_knn_vc | 8.7% | 25.0% | 32.9% | 33.5% |
189
+ | orpheus_tts | 9.1% | 24.0% | 33.6% | 33.3% |
190
+ | sesame_tts | 9.1% | 24.0% | 33.6% | 33.3% |
191
+ | orpheus_knn_vc | 9.1% | 24.0% | 33.6% | 33.3% |
192
+ | sesame_knn_vc | 9.1% | 24.0% | 33.6% | 33.3% |