72 Books & arts The EconomistSeptember 5th 2020
21Boys—desperate to win fame, status and
money from rapping. He shows how their
musical and lyrical talent is only a minor
part of what determines success. As impor-
tant are attempts to win a reputation on-
line as authentic gangsters, despite their
sometimes feeble efforts to acquire weap-
ons, cash or other props essential for build-
ing credibility.
Aspiring stars must at least pretend
they are heavily involved in conflict. “It is a
hyper-violent context,” says Mr Stuart;
even conducting research was risky at
times. A young rapper mishandling a gunnearly shot him. He describes learning how
the Corner Boys lurked on the social-media
feed of a rival driller, here called Smoky-P,
fond of posting selfies. They identified a
shop where he regularly bought alcohol,
drove by and fired at him several times, ap-
parently as he prepared to take a photo-
graph of himself.
The author is reluctant to simply blame
his subjects for such appalling acts. For
many characters he uses pseudonyms to
avoid stirring more conflict, or inviting at-
tention from police. Most such young men,
he notes, are traumatised by street shoot-ings they have experienced from early
childhood. Broken families produce ill-
educated men who go on to choose be-
tween menial service jobs or drug dealing.
Youngsters have few role models to follow.
In slums, favelas and shantytowns else-
where, ambitious people dream of getting
out by playing football. In Chicago drill, de-
spite its gang ties, serves that end. For a few
rappers who earn enough online notoriety,
such as Chief Keef, record contracts and
wealth follow. More interesting is the fate
of larger numbers of smaller fry like the
Corner Boys. Their relatively modest musi-Johnson Lost in dictation
Is it possible to “write” using speech-to-text software?V
oice technologyhas come a long
way. Just a few years ago, it would’ve
been unusable, so much so that many
people gave up on it in the early years.
But now, those who follow the tech-
nology know that it has gotten consid-
erably better over the last five to ten
years. Truly effective voice technology
promises many good things. Not only
does it allow you to command devices
like smartphones or use wired home
devices without the use of hands, but
also many other applications. For ex-
ample, you can identify someone by
their voice if you need to know who they
are over the phone. Your bank will find
this very useful.
What about writing with your voice?
This could be a great boon for people who
lack mobility for one reason or another.
But not only that. Busy people who type
badly, and find it easier to talk, might
find it much easier to dictate their mes-
sages than they would find to sit and type
them. Everyone remembers the bosses in
old movies who shout things like quo-
tation mark Ms Johnson! Take a memo.
Rotation marked.
But writing with your voice raises
several interesting questions. How easy
is it actually? Human speech involves a
lot more starting and stopping with
errors and the need for repairing man-
gling sentences than you may think.
Writing may be an unnatural act, but
once it is learned, the first draft of a piece
of writing is a lot more usable than the
“first draft” of a bit of speech. Anyone
tempted to doubt this proposition might
try listening to an interview, even one
with a highly articulate speaker, and
transcribing every single word that that
person says. It will quickly become obvi-
ous that even gifted speakers make lots
of mistakes. That’s not usually a problemfor structure only. It has been edited for
length, with all of the original errors kept
in. Though this paragraph two is being
dictated, which means I have no idea
how this is turning out, here are guesses
about how this will read.
The first guess is that the literal accu-
racy of the dictation software will be
extremely high. In other words there
won’t be many cases where the software
has heard one word incorrectly and
transcribed another. But the other guess
is that the readability of this column will
be rather bad.
In other words, the blame is not with
the technology, which turns out to be
rather good. Speaking into writing relies
on a much better human brain than the
one we currently possess. Writing is
hard. There’s a reason it can’t be done at
the speed of speech, in real time.
To elucidate—and this was written
after the fact, rather than dictated—
paragraph breaks were added after tran-
scription. Punctuation had to be spoken
aloud, but after a full stop the first word
in the new sentence was capitalised
automatically. Some minor punctuation
marks were added to improve clarity.
Only a handful of words out of almost
800 were transcribed incorrectly, among
them “rotation marked” (“quotation
mark”), “two” (“too”) and “it” (“if”). To
improve accuracy, your author “trained”
the software, Dragon Dictate, for a few
seconds beforehand, reading a pre-
written passage aloud. An external mi-
crophone of reasonable quality also
helped a great deal.
Nonetheless, The Economist’s style
mavens have their heads in their hands
at the ugly Americanisms and egregious
stylistic infractions. It will be a relief to
them, and to the reader, that Johnson will
not be dictating any future columns.when we talk to each other, but it makes
for extraordinarily messy looking writing.
To make turning the spoken word into
coherent writing requires lots of planning.
You’ll need some kind of notes or other
organiser to make it work, which only
brings us back to the original problem.
Those who need to write with their voice
will first need to write a structure to then
write from. If you are unable to write, this
does not solve the problem.
Another question turning speech into
writing raises is one of style. How would
writing change it more people spoke their
writing rather than typing? Chances are
that at the very least they would come up
with many more short sentences and more
concrete language, which is good. But they
would probably also rely on pre-assem-
bled phrases and clichés a lot more often,
which would definitely be a bad thing.
Good writing requires slow thought,
which is not available when you are speak-
ing at full clip.
To test this proposition this column
has been not written but dictated. It was
composed from brief notes written down