How to apply zero‐shot learning to text data in substance use research: An overview and tutorial with media data
Addiction,
Год журнала:
2024,
Номер
119(5), С. 951 - 959
Опубликована: Янв. 11, 2024
Abstract
A
vast
amount
of
media‐related
text
data
is
generated
daily
in
the
form
social
media
posts,
news
stories
or
academic
articles.
These
provide
opportunities
for
researchers
to
analyse
and
understand
how
substance‐related
issues
are
being
discussed.
The
main
methods
large
(content
analyses
specifically
trained
deep‐learning
models)
require
substantial
manual
annotation
resources.
machine‐learning
approach
called
‘zero‐shot
learning’
may
be
quicker,
more
flexible
fewer
Zero‐shot
learning
uses
models
on
large,
unlabelled
(or
weakly
labelled)
sets
classify
previously
unseen
into
categories
which
model
has
not
been
trained.
This
means
that
a
pre‐existing
zero‐shot
can
used
without
need
task‐specific
training.
particularly
important
analysing
time
critical.
article
describes
relatively
new
concept
it
applied
substance
use
research,
including
brief
practical
tutorial.
Язык: Английский
Research Letter: Unveiling the boundaries: Analysing advertising breaches in the promotion of medicinal cannabis in Australia
Australian & New Zealand Journal of Psychiatry,
Год журнала:
2025,
Номер
unknown
Опубликована: Янв. 10, 2025
Язык: Английский
What is the hype on #MedicinalCannabis in the United States? A content analysis of medicinal cannabis tweets
Drug and Alcohol Review,
Год журнала:
2023,
Номер
43(1), С. 28 - 35
Опубликована: Фев. 21, 2023
Abstract
Introduction
Medicinal
cannabis
is
now
legal
in
44
US
jurisdictions.
Between
2020
and
2021
alone,
four
jurisdictions
legalised
medicinal
cannabis.
The
aim
of
this
study
to
identify
themes
tweets
from
with
different
statuses
January
June
2021.
Methods
A
total
25,099
historical
51
were
collected
using
Python.
Content
analysis
was
performed
on
a
random
sample
accounting
for
the
population
size
each
(
n
=
750).
Results
presented
separately
by
posted
where
all
use
(non‐medicinal
medicinal)
‘fully
legalised’,
‘illegal’
‘medical‐only’
use.
Four
identified:
‘Policy’,
‘Therapeutic
value’,
‘Sales
industry
opportunities’
‘Adverse
effects’.
Most
public.
most
common
theme
related
‘Policy’
(32.5%–61.5%
tweets).
Tweets
value’
prevalent
accounted
23.8%–32.1%
tweets.
Sales
promotional
activities
prominent
even
illegal
(12.1%–26.5%
Fewer
than
10%
about
intoxication
withdrawal
symptoms.
Discussion
Conclusion
This
has
explored
if
content
differed
status.
pro‐cannabis
they
policy,
therapeutic
value,
sales
opportunities.
unsubstantiated
health
claims,
adverse
effects
crime
warrants
continued
surveillance
as
these
conversations
could
allow
us
estimate
cannabis‐related
harms
inform
surveillance.
Язык: Английский
Editorial: AI and new digital technologies have transformed alcohol and other drug industries lobbying
Drugs Habits and Social Policy,
Год журнала:
2024,
Номер
25(1), С. 1 - 5
Опубликована: Май 10, 2024
niveau
recherche,
publiés
ou
non,
émanant
des
établissements
d'enseignement
et
de
recherche
français
étrangers,
laboratoires
publics
privés.
The Intersection of Rurality, Unmet Treatment Need, and Cannabis Use for Adults with Chronic Pain
Cannabis,
Год журнала:
2024,
Номер
unknown
Опубликована: Май 16, 2024
Objective:
Rates
of
chronic
pain
are
higher
among
rural
versus
urban
individuals
and
experience
levels
socioeconomic
disadvantage,
poor
or
no
health
insurance
coverage,
unmet
treatment
need.
Medical
cannabis
is
legal
in
Oklahoma.
With
40%
Oklahoma’s
population
living
areas,
nearly
17%
uninsured,
the
medical
legalization
may
present
as
an
accessible
relatively
low-cost
alternative
treatment,
particularly
for
those
with
pain.
This
study
investigated
differences
use
by
(vs.
urban)
status
met)
need
adults
without
Method:
To
be
eligible,
participants
had
to
18
years
older,
reside
Oklahoma,
able
read
write
English-language
surveys.
Results:
The
sample
(N
=
3622)
was
primarily
made
up
non-Hispanic
White
(70.4%)
females
(53.8%)
their
early
middle
age
(M
41.80,
SD
16.88),
employed
full-time
part-time
(53.8%),
some
college/technical
school
(37.2%)
a
bachelor’s
degree
(28.5%).
Nearly
one-fifth
(18.2%)
endorsed
pain,
were
eight
times
more
likely
report
past
30-day
use.
No
difference
detected
when
only
rurality
residence)
examined.
Among
who
dwelling
reported
almost
two
use,
compared
Conclusions:
In
areas
benefit
from
increasing
access
well
education
on
harm
reduction
strategies
inform
healthcare
decision-making.
Язык: Английский