Lakoc commited on
Commit
989b69f
·
verified ·
1 Parent(s): bd2e2ab

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +30 -0
README.md CHANGED
@@ -68,6 +68,36 @@ For detailed inference and full pipelines, refer to:
68
  👉 [DiCoW GitHub inference repo](https://github.com/BUTSpeechFIT/DiCoW)
69
 
70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
71
 
72
  ## Citation
73
 
 
68
  👉 [DiCoW GitHub inference repo](https://github.com/BUTSpeechFIT/DiCoW)
69
 
70
 
71
+ ### tcpWER/CER (%) on the MLC-SLM development set
72
+
73
+ | Language | Baseline (GT) | DiCoW (GT) | FT (GT) | Baseline (Real diar) | DiCoW (Real diar) | FT (Real diar) |
74
+ |----------------|---------------|------------|---------|-----------------------|-------------------|----------------|
75
+ | American En. | 14.1 | 20.6 | 11.1 | 53.7 | 36.5 | 22.5 |
76
+ | Australian En. | 11.7 | 19.4 | 7.4 | 52.6 | 23.6 | 13.0 |
77
+ | British En. | 10.1 | 16.7 | 7.7 | 71.9 | 26.1 | 17.6 |
78
+ | Filipino En. | 9.2 | 17.7 | 7.5 | 50.4 | 25.5 | 15.2 |
79
+ | Indian En. | 14.0 | 14.3 | 13.3 | 70.7 | 14.9 | 14.0 |
80
+ | French | 28.1 | 27.7 | 16.1 | 96.0 | 37.8 | 27.5 |
81
+ | German | 20.7 | 21.2 | 23.9 | 86.7 | 30.1 | 27.3 |
82
+ | Italian | 17.9 | 16.2 | 12.3 | 83.3 | 19.8 | 16.4 |
83
+ | Japanese (\*) | 21.6 | 19.2 | 13.7 | 71.3 | 25.8 | 23.3 |
84
+ | Korean (\*) | 13.8 | 12.8 | 8.5 | 59.6 | 24.5 | 22.8 |
85
+ | Portuguese | 21.2 | 24.5 | 19.5 | 118.8 | 33.1 | 29.7 |
86
+ | Russian | 17.7 | 17.6 | 11.6 | 69.2 | 22.5 | 16.7 |
87
+ | Spanish | 12.3 | 11.6 | 8.7 | 75.6 | 18.2 | 16.3 |
88
+ | Thai (\*) | 14.5 | 31.9 | 14.2 | 83.6 | 34.4 | 20.1 |
89
+ | Vietnamese | 27.2 | 30.0 | 15.3 | 82.8 | 33.8 | 24.7 |
90
+ | **Overall** | **16.8** | **22.0** | **12.9**| **76.1** | **28.4** | **20.8** |
91
+
92
+ > *Results marked with an asterisk (*) are reported using tcpCER, following the official evaluation protocol.*
93
+
94
+ **Notes:**
95
+
96
+ - GT = Ground-Truth Segmentation
97
+ - Real diar = Real Diarization
98
+ - Baseline uses Whisper large-v3 with chunked inference + finetunned Pyannote diarization.
99
+ - DiCoW uses fine-tuned DiariZen diarization.
100
+
101
 
102
  ## Citation
103