Swiss-German-Speech-Data---Meeting-Notes.md

+## Meeting April 29, 2026
+
+### Participants
+
+Yixuan Xu, Daniel Perruchoud, Roberto Salomone, Michael Graber
+
+### Previous Action Items
+
+* [ ] Propose next steps for SFT data
+* [ ] Roberto: Licence agreement, also partially for Regionaljournal & Espresso broadcasts (needs to be commercially usable and openly shareable)
+* [x] FHNW: propose new, 2k test split for SPC_R
+* [x] FHNW: WER for FHNW transcription model
+
+### Discussion Points
+
+* Yixuan shares latest evalution results with respect to ablation study of SSR data
+  * SSR data lowers WER/CER \~20%
+  * More data would help even more!
+  * However, we should find more human labelled transcripts (only working with pseudo-labelled data is considered problematic)
+    * Maybe archive data would be useful 
+
+### New Action Points
+
+* [ ] FHNW: Analysis of SwissDial evaluation
+
 ## Meeting April 22, 2026
 
 ### Participants
@@ -24,7 +49,7 @@ Yixuan Xu, Fabian Jeitziner, Roberto Salomone, Michael Graber
 ### New Action Points
 
 * [ ] FHNW: propose new, 2k test split for SPC_R
-* [ ] FHNW: WER for FHNW transcription model 
+* [ ] FHNW: WER for FHNW transcription model
 
 ## Meeting April 15, 2026