You will need the switchboard audio files from https://catalog.ldc.upenn.edu/ldc97s62. These are sadly not openly available, but your institution might have an LDC license.
If you are running on the ISL cluster:
ln -s /project/earsData/swbLinks adc
The ISL utterance database (should not be needed anymore, replaced by the ISIP transcriptions)
ln -s /project/ears2/db/train/ db
The original transcriptions are from https://www.isip.piconepress.com/projects/switchboard/
scripts/getdata.sh
You can compare the integrity of the data files with the ones I used by running
sha256sum -c sha256sums.txt
The utterance_is_backchannel.json contains a dump of backchannel vs non-backchannel data for all utterances in Switchboard, generated by python -m extract.standalone_save_backchannels_to_json.