Upload folder using huggingface_hub

README.md

@@ -42,6 +42,7 @@ Models used:
 
 An 80Hz highpass filter was applied to the source audio prior to segmentation to reduce incidence of the recording environment's low frequency noise.
 
+TODO: update numbers + row count above (bumped to 800ms silence from 350)
 In total, 26.07 hours of audio remains, as 25091 conversational turns. Some transcription errors are to be expected.
 
 # The Expresso Dataset

train.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4041b06efd24066f8438188b0324449343bd214fa50898dc86f2bdefee74a2cc
-size 13504559003
+oid sha256:015f05560074dc50456ccd7d3b5e2b571c242fea23ed3fce2150421bcd295c91
+size 14158312562