#4 benchmark youtube auto transcript

0:04
okay
0:05
which data set would you use to
0:07
benchmark performance
0:10
uh
0:12
well i i'd probably use library speech
0:14
library speech is something that tons of
0:16
people publish on so it's always easy to
0:18
look up what the baseline result is
0:20
uh
0:21
yeah certainly in calder speech is the
0:23
most convenient baseline to use but
0:26
people use it in many other toolkits as
0:28
well
0:30
if someone were to publish a paper
0:32
which data set would you recommend for
0:34
them to show the results
0:36
yeah i think yeah especially for papers
0:38
liberty speech is a good one for uh
0:42
for experiments
0:45
i mostly tend to use library speech but
0:47
sometimes uh
0:49
sometimes other data sets like uh
0:53
let me think
0:54
i mean switchboard is a common one but
0:56
switchboard is not free
0:59
yeah
1:01
thank you
1:03
bye

”Especially for papers liberty speech is a good one.”

“Switchboard is a common one but switchboard is not free.”