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

@@ -2,6 +2,34 @@
 
 ### Participants
 
+Yixuan Xu, Roberto Salomone, Daniel Perruchoud, Michael Graber
+
+### Previous Action Items
+
+- [ ] Roberto: Confirm huggingface dump for Espresso, Regionaljournal Licence for hugging face cc by 4.0 nc (?)
+- [x] Yixuan: discuss with Imanol how long srg broadcasts should be chunked, if at all
+- [x] Michael : write email with precise license request for the shared Regionaljournal & Espresso data to Roberto
+
+### Discussion Items
+
+* Apertus vision model develops fine
+* There is a window of opportunity for getting data into pre-training
+* Transcript evaluation app ready this afternoon, preview here: https://speechmodeleval.informatik.fhnw.ch/
+* Roberto: Licences, what use cases? Due date? (latest Monday, Feb 16)
+* VAD support might be helpful
+* 173 M tokens from _Gemeinderat Zürich Dateset_ generated
+* phase-2 interleaving sequences should be 5-10 s
+* Speech synthesis has limitations, since it would require a variety of speakers, genders, timbres etc.
+* GLM team will be at AI Center, could join for talk tomorrow Thursday at 12pm, MBZUAI on Friday
+* Roberto: when will be the Apertus 1.5 be ready? -\> Yixuan, pre-training \~ end of March
+
+### New Action Items
+
+* [ ] Share code for interleaving preprocessing
+* [ ] Identify FHNW students to work in the cluster with Yixuan
+
+### Participants
+
 Yixuan Xu, Roberto Salomone, Michael Graber
 
 ### Previous Action Items