ListLinePlot[LowpassFilter[With[{dataset = Dataset[{<|"" -> 0,
"text" ->
"Science has gotten big; how does quality control usually \
work in some big system? Analog is", "start" -> 0.0,
"duration" -> 91|>, <|"" -> 1,
"text" ->
"regulation in markets, it's easier to make safe food than \
unsafe food, it's easier to make bad ", "start" -> 91,
"duration" -> 186|>, <|"" -> 2,
"text" ->
"science than good science. You have peers, you have people \
who can give you input and if you're ", "start" -> 186,
"duration" -> 282|>, <|"" -> 3,
"text" ->
"rational and actually want to create good scientific \
progress, then you talk to them... you're not ", "start" -> 282,
"duration" -> 389|>, <|"" -> 4,
"text" ->
"regulated to do peer review but if you mess up and start to \
release garbage there's some way ", "start" -> 389,
"duration" -> 482|>, <|"" -> 5,
"text" ->
"people can find out about it and that'll affect your \
ability to spread these ideas... and probably ", "start" -> 482,
"duration" -> 589|>, <|"" -> 6,
"text" ->
"they'll eventually crash. Maybe one can never escape the \
feral cats in the case that it's still... a", "start" -> 589,
"duration" -> 686|>, <|"" -> 7,
"text" ->
"case in which there's higher variance; people who are doing \
crazy stuff and getting high velocity ", "start" -> 686,
"duration" -> 784|>, <|"" -> 8,
"text" ->
"doing crazy stuff even though it's crazy and yes I can see \
that can happen. The one thing that ", "start" -> 784,
"duration" -> 879|>, <|"" -> 9,
"text" ->
"hasn't happened and that journals could do is computational \
publication, they don't have to do ", "start" -> 879,
"duration" -> 974|>, <|"" -> 10,
"text" ->
"typesetting or copyediting but what they could do is get \
code to actually work, the code that they ", "start" -> 974,
"duration" -> 1073|>, <|"" -> 11,
"text" -> "publish actually works.", "start" -> 9.92,
"duration" -> 1096|>}]},
CosineDistance[#[[1]], #[[2]]] & /@
Partition[
Mean[NetModel[
"BERT Trained on BookCorpus and Wikipedia Data"][#]] & /@
StringRiffle /@
Select[Normal[
dataset[Function[{startTime, endTime},
Select[#start >= startTime && #start <
endTime &]][#[[1]], #[[2]]], "text"] & /@
Partition[
Range[1,
Ceiling[dataset[-1, "start"] +
dataset[-1, "duration"]], .6*0.5], 2, 1]],
UnsameQ[#, {}] &], 2, 1]], 0.5], MeshStyle -> Red,
Mesh -> All]
Oh the Kindle version is out...come up from antiquity and see the Linguistic Relations & Proof Spaced Topology.

E-books...isn't that the spirit to see a linguistic diagram of how the words flow? You could do the Centennial.