1
00:00:01,380 --> 00:00:04,260
Unknown: Hello, and welcome to
this workshop. This is Kristi

2
00:00:04,260 --> 00:00:07,830
Thompson. I'm a research data
management library and on the

3
00:00:07,830 --> 00:00:10,020
research and scholarly
communications team in the

4
00:00:10,020 --> 00:00:15,450
library. Basically I help people
access, do research with manage

5
00:00:15,480 --> 00:00:18,360
and archive all kinds of data.
I've been asked to provide a

6
00:00:18,360 --> 00:00:22,050
pretty brief discussion,
introduction to how to access

7
00:00:22,050 --> 00:00:26,370
and do research with social
media data. This isn't going to

8
00:00:26,370 --> 00:00:29,790
be an exhaustive overview, I'm
hoping instead to spark some

9
00:00:29,790 --> 00:00:34,080
ideas about the possibilities of
social media research. And I

10
00:00:34,080 --> 00:00:36,570
very much hope that if any of
you would like to explore this

11
00:00:36,570 --> 00:00:40,140
further, you'll reach out to me
and my colleagues, we can help

12
00:00:40,140 --> 00:00:40,230
you

13
00:00:40,260 --> 00:00:40,860
get going.

14
00:00:44,890 --> 00:00:48,790
Social media data is data
derived by collecting material

15
00:00:48,790 --> 00:00:52,000
that people share on social
networking sites and assembling

16
00:00:52,000 --> 00:00:56,680
it into a data set for analysis.
Social media data consists of

17
00:00:56,680 --> 00:01:01,570
content, which is the data that
words images, videos and

18
00:01:01,570 --> 00:01:06,280
reactions, such as thumbs up
shared by users of social media

19
00:01:06,280 --> 00:01:11,740
platforms. And social media data
also includes metadata, which is

20
00:01:12,250 --> 00:01:14,980
information about the data and
about the author of the data.

21
00:01:16,480 --> 00:01:19,990
metadata is also an important
part of social media related

22
00:01:19,990 --> 00:01:24,100
research. It includes platform
generated metadata, which is

23
00:01:24,100 --> 00:01:26,380
linked to an embedded
information that describes

24
00:01:26,380 --> 00:01:30,370
things like when the content was
shared, who shared the content,

25
00:01:30,610 --> 00:01:35,680
how often it was reshared, and
where the user is from. And this

26
00:01:35,680 --> 00:01:39,850
also includes the user generated
metadata, such as on Twitter

27
00:01:39,850 --> 00:01:47,020
things like hashtags and
mentions. So that's the what.

28
00:01:47,590 --> 00:01:52,090
Now, why do people analyze
social media data? It turns out

29
00:01:52,090 --> 00:01:55,180
social media data is great for
social research because social

30
00:01:55,180 --> 00:01:59,350
media allows access to the
unprompted unscripted opinions

31
00:01:59,380 --> 00:02:03,880
of users in their own words.
Now, traditional research

32
00:02:03,910 --> 00:02:07,900
usually elicits user opinions
under artificial continued

33
00:02:08,260 --> 00:02:13,150
conditions, such as surveys or
focus groups. In contrast,

34
00:02:13,150 --> 00:02:16,120
social media to discover
reactions to events as they

35
00:02:16,120 --> 00:02:21,190
happen in real time without
prompting, without any observer

36
00:02:21,190 --> 00:02:25,210
effect from the researcher. And
social media allows you to also

37
00:02:25,210 --> 00:02:29,320
analyze connections between
people, and trace how ideas

38
00:02:29,320 --> 00:02:34,540
spread. Twitter is particularly
used, suited for this type of

39
00:02:34,540 --> 00:02:37,210
research because of data
availability, wide use of

40
00:02:37,210 --> 00:02:41,470
hashtags and easily traced back
works. But I'll be talking about

41
00:02:41,470 --> 00:02:44,050
some of the other types of
social media you can do research

42
00:02:44,050 --> 00:02:48,340
with as well. They'll have their
unique advantages and

43
00:02:48,340 --> 00:02:54,160
disadvantages. There are also
reasons why social media data

44
00:02:54,160 --> 00:02:57,190
needs to be used carefully, and
why results from Social Media

45
00:02:57,190 --> 00:03:01,630
Research need to be treated with
caution. Social media is

46
00:03:01,630 --> 00:03:05,740
performative, both positive and
negative, online feelings are

47
00:03:05,740 --> 00:03:09,100
often overstated. And interest
in a topic may not actually

48
00:03:09,100 --> 00:03:13,630
translate into further actions.
We've all heard the stereotype

49
00:03:13,630 --> 00:03:18,730
of the social media warrior who
talks about things a lot, but

50
00:03:19,060 --> 00:03:24,370
doesn't do much. You In
addition, users of social media

51
00:03:24,460 --> 00:03:28,570
are not representative of
populations. So there's self

52
00:03:28,570 --> 00:03:31,810
selection bias. Some people
produce far more output than

53
00:03:31,810 --> 00:03:36,010
others, which creates quite a
bit of bias. You basically end

54
00:03:36,010 --> 00:03:40,030
up with a lot of data on people
who like to talk on social media

55
00:03:40,030 --> 00:03:45,700
a lot. And then there's also
just difficulty created by the

56
00:03:45,700 --> 00:03:49,900
scope of social media data. Most
searches will retrieve large

57
00:03:49,900 --> 00:03:53,260
amounts of content that is
difficult to come through with

58
00:03:53,260 --> 00:03:55,570
much of the data being
irrelevant to a research

59
00:03:55,630 --> 00:04:02,200
question, or impossible to
interpret. Basically, you always

60
00:04:02,200 --> 00:04:05,050
need to think very carefully
about what exactly you're

61
00:04:05,050 --> 00:04:09,970
studying. Who are the people
that use this social media? What

62
00:04:09,970 --> 00:04:12,790
are their demographic
characteristics, who's being

63
00:04:12,790 --> 00:04:15,790
left out of this sample? Is
there another way to get this

64
00:04:15,790 --> 00:04:16,510
information?

65
00:04:18,010 --> 00:04:19,300
Just like all research.

66
00:04:22,350 --> 00:04:25,980
There are quite a lot of social
media platforms out there. Some

67
00:04:25,980 --> 00:04:29,520
of them are easier to do
research on than others. Twitter

68
00:04:29,520 --> 00:04:33,000
is probably the most popular and
easiest to use social media

69
00:04:33,000 --> 00:04:36,660
network research. tweets are
popular PR analysis because

70
00:04:36,660 --> 00:04:41,250
they're short, consistent text,
and contain user generated

71
00:04:41,250 --> 00:04:44,790
metadata, which makes them
suitable for automated analysis

72
00:04:44,790 --> 00:04:48,960
and collection. Also, Twitter
actually allows people to access

73
00:04:48,960 --> 00:04:52,140
Twitter data through the Twitter
API, or application programming

74
00:04:52,140 --> 00:04:55,800
interface, which makes it easy
to get data automatically and

75
00:04:55,800 --> 00:04:59,460
import it into a computer
program. We'll be talking later

76
00:04:59,460 --> 00:05:04,080
about how Take advantage of
API's to harvest data. Even if

77
00:05:04,080 --> 00:05:06,660
you aren't very technical, it's
actually pretty easy to do.

78
00:05:09,630 --> 00:05:12,930
YouTube is widely considered
both a video sharing platform

79
00:05:12,930 --> 00:05:16,260
and a social media network due
to its quantity of user

80
00:05:16,260 --> 00:05:21,180
generated content. Both the
videos and the comments and

81
00:05:21,180 --> 00:05:23,910
discussion that take place on
the videos are available for

82
00:05:23,910 --> 00:05:28,650
analysis. video analysis can be
pretty time consuming, but

83
00:05:28,650 --> 00:05:31,860
content analysis can be quite
rich and revealing. And

84
00:05:31,860 --> 00:05:35,250
harvesting comments for a single
video we're often produce a data

85
00:05:35,250 --> 00:05:39,690
set that's a good size for basic
analysis. YouTube also has an

86
00:05:39,690 --> 00:05:46,800
API for harvesting data. online
news and discussion forums such

87
00:05:46,800 --> 00:05:50,580
as Reddit allow the analysts to
collect longer form discourse,

88
00:05:51,030 --> 00:05:55,110
you can collect the text of a
single long discussion using

89
00:05:55,110 --> 00:05:59,730
copy and paste from your browser
or read it also, again has an

90
00:05:59,730 --> 00:06:06,450
API for automated searching of
data. Some partial popular

91
00:06:06,450 --> 00:06:09,780
social media networks such as
Instagram or Facebook are less

92
00:06:09,780 --> 00:06:12,810
suitable for analysis because
the terms of service do not

93
00:06:12,810 --> 00:06:16,200
allow data collection or because
you can only access limited

94
00:06:16,200 --> 00:06:20,760
information due to people's
privacy settings. Facebook is

95
00:06:20,760 --> 00:06:26,310
one limited social media network
that does make some specific

96
00:06:26,310 --> 00:06:30,030
data sets available to
researchers. But in general,

97
00:06:30,030 --> 00:06:36,240
their data is not available for
scraping. You need to be

98
00:06:36,240 --> 00:06:39,900
especially cautious about using
some platforms, particularly if

99
00:06:39,900 --> 00:06:43,470
you're concerned about being
sued. LinkedIn is an interesting

100
00:06:43,470 --> 00:06:47,250
case, their terms of service
explicitly disallows scraping,

101
00:06:47,340 --> 00:06:50,670
and they have accepted attempted
to sue outside entities that

102
00:06:50,670 --> 00:06:55,710
scraped their data, scraping
LinkedIn public information, but

103
00:06:55,710 --> 00:06:59,970
you can see well not logged in,
was however the rules legal

104
00:07:00,000 --> 00:07:05,430
following a court case. However,
Terms of Service still disallow

105
00:07:05,460 --> 00:07:10,560
automatic access, and they may
attempt to block any attempt to

106
00:07:10,560 --> 00:07:15,480
access their data automatically.
Instagrams another one, their

107
00:07:15,480 --> 00:07:19,050
terms of service do not allow
scraping the data apart from the

108
00:07:19,050 --> 00:07:23,340
crowdtangle public data, which
is available from the books data

109
00:07:23,340 --> 00:07:28,860
for public good intercept face.
And their privacy settings on

110
00:07:28,860 --> 00:07:31,860
Instagram make it quite
difficult to get at the data.

111
00:07:34,590 --> 00:07:38,460
Google Groups is another one.
This isn't as popular social

112
00:07:38,460 --> 00:07:41,850
platform as some of the others,
but I'll explain why it's worth

113
00:07:41,850 --> 00:07:46,140
considering. Now their terms of
service describe scraping of

114
00:07:46,140 --> 00:07:50,790
content as spam, and it may
result in deletion of your

115
00:07:50,790 --> 00:07:54,000
account. However, on the other
hand, they don't explicitly make

116
00:07:54,000 --> 00:07:58,500
any reference to suing people.
Some Google Groups are public

117
00:07:58,500 --> 00:08:02,640
and can be accessed without
account and account. These may

118
00:08:02,640 --> 00:08:06,450
be legitimate targets for
research. These include the

119
00:08:06,450 --> 00:08:12,630
Usenet archive, dating back to
1981, an archive of online

120
00:08:12,630 --> 00:08:14,940
discussion that took place
before the web was even

121
00:08:14,940 --> 00:08:19,290
developed. So if you're doing
structural research, looking

122
00:08:19,290 --> 00:08:23,850
into Google Groups may be worth
the effort. For any other

123
00:08:23,850 --> 00:08:28,020
platforms, check the terms of
service to see what's allowed,

124
00:08:29,160 --> 00:08:30,990
and how you can go about getting
access to it.

125
00:08:36,450 --> 00:08:39,300
If you need to sign up for an
account on a platform in order

126
00:08:39,300 --> 00:08:42,840
to see the data, check the terms
of service to find out exactly

127
00:08:42,840 --> 00:08:46,950
what you're agreeing to. If data
can be accessed and read without

128
00:08:46,950 --> 00:08:50,730
an account, you're probably safe
legally when it comes to doing

129
00:08:50,730 --> 00:08:54,510
research using it. But the
platform provider may well have

130
00:08:54,510 --> 00:08:59,160
made it technically difficult.
There's been some case law on

131
00:08:59,160 --> 00:09:02,790
this. And the US as I've
mentioned LinkedIn data scraping

132
00:09:02,790 --> 00:09:07,500
was ruled legal. However, in
Canada, a particular federal

133
00:09:07,500 --> 00:09:11,850
court case ruled web scraping
illegal, at least if done

134
00:09:11,850 --> 00:09:16,050
deceptively for profit. There
hasn't been a ruling on more

135
00:09:16,050 --> 00:09:20,640
general cases such as accessing
data and automatically is for

136
00:09:20,640 --> 00:09:27,750
research purposes. In general,
scraping social media data is

137
00:09:27,750 --> 00:09:31,320
legally safe if you comply with
the terms of service and stick

138
00:09:31,320 --> 00:09:36,720
with data that's available to
the general public. So that's

139
00:09:36,720 --> 00:09:41,310
legality. How about ethics? I
reached out to representatives

140
00:09:41,310 --> 00:09:43,920
of westerns research ethics
board while preparing the

141
00:09:43,920 --> 00:09:46,920
slides, make sure that I was
complying with local practice

142
00:09:48,240 --> 00:09:53,880
under tcps to Article 2.2
publicly available information

143
00:09:53,910 --> 00:09:58,350
is exempt from research ethics
board review. This includes

144
00:09:58,350 --> 00:10:01,950
identifiable information, so
Social media data where there is

145
00:10:01,950 --> 00:10:06,570
no expectation of privacy. So
you don't even know need to go

146
00:10:06,570 --> 00:10:11,070
through review for this type of
research. Basically, what it

147
00:10:11,070 --> 00:10:15,240
comes down to is naturalistic
observation of public behavior

148
00:10:15,240 --> 00:10:19,020
does not require review. Even if
the observation is occurring

149
00:10:19,020 --> 00:10:23,130
online. You're just watching
people do what they normally do.

150
00:10:23,880 --> 00:10:26,880
This is distinct from
information in private groups or

151
00:10:26,880 --> 00:10:31,080
chats. Every group is available
to all users of a platform and

152
00:10:31,080 --> 00:10:34,800
anyone is allowed to register by
the platform, the information

153
00:10:34,800 --> 00:10:38,490
may be considered public. If you
need to ask permission from a

154
00:10:38,490 --> 00:10:42,390
moderator or other entity to
join a group. Participants in

155
00:10:42,390 --> 00:10:45,660
that group may have a reasonable
expectation of privacy, and you

156
00:10:45,660 --> 00:10:48,720
may have to gain their consent
in order to do research on them.

157
00:10:49,860 --> 00:10:53,580
Also, if the researcher is doing
any sort of an intervention,

158
00:10:54,510 --> 00:10:57,570
asking questions or
participating in conversation in

159
00:10:57,570 --> 00:11:01,440
an online group, then review and
possibly informed consent will

160
00:11:01,440 --> 00:11:08,310
be required. Like to be focusing
on Twitter data over the next

161
00:11:08,310 --> 00:11:11,730
few slides. Twitter is the most
popular data for Social Media

162
00:11:11,730 --> 00:11:14,850
Research because of its relative
ease of access, and the wide use

163
00:11:14,850 --> 00:11:21,930
of hashtags that make it easy to
track how ideas spread. Until

164
00:11:21,930 --> 00:11:24,930
recently, the only options
practicing Twitter data were the

165
00:11:24,930 --> 00:11:29,430
free standard search, which
severely limits the amount of

166
00:11:29,430 --> 00:11:32,970
data, you could access to only
the most recent few days or

167
00:11:32,970 --> 00:11:38,160
weeks, and expensive premium or
Enterprise Search. These are

168
00:11:38,160 --> 00:11:42,330
both through the Twitter API.
However, Twitter recently added

169
00:11:42,330 --> 00:11:46,860
the new free academic research
track, which allows access to

170
00:11:46,860 --> 00:11:52,230
the full historical corpus of
data. Whoever access to this

171
00:11:52,230 --> 00:11:55,920
stream requires application and
approval of your application by

172
00:11:55,920 --> 00:11:59,580
someone at twitter. I actually
don't know how long this

173
00:11:59,580 --> 00:12:02,760
qualifies or what sort of
factors they consider in making

174
00:12:02,760 --> 00:12:07,050
approval because it's clear from
the terms that I don't qualify

175
00:12:07,050 --> 00:12:11,040
at all, so I couldn't even
apply. To apply you must be

176
00:12:11,070 --> 00:12:14,640
employed as a researcher at an
academic institution, or a

177
00:12:14,640 --> 00:12:18,510
graduate student working on a
thesis or dissertation, not

178
00:12:18,510 --> 00:12:23,370
class project and not an
undergrad. And you must have a

179
00:12:23,370 --> 00:12:29,850
specific, clearly defined
research project. For everyone

180
00:12:29,850 --> 00:12:33,600
else, standard searches still
available. Standard search

181
00:12:33,600 --> 00:12:37,200
through the API allows access to
the most recent seven to nine

182
00:12:37,200 --> 00:12:43,470
days of data. Another option is
using existing archive data

183
00:12:43,470 --> 00:12:47,880
collections of tweets, you can
go out and search for a topic

184
00:12:47,880 --> 00:12:52,950
such as COVID-19, Twitter, and
data or corpus to find existing

185
00:12:52,950 --> 00:12:57,540
collections, you might get
lucky. For example, George

186
00:12:57,540 --> 00:13:00,420
Washington University is
assembled a variety of large

187
00:13:00,420 --> 00:13:03,510
collections of current and
historical Twitter data on

188
00:13:03,510 --> 00:13:07,050
various topics, and provide some
instructions on how to use them.

189
00:13:07,770 --> 00:13:10,410
Many other Twitter data archives
are available including a

190
00:13:10,410 --> 00:13:17,490
massive unsorted collection at
the Internet Archive. The

191
00:13:17,490 --> 00:13:20,910
easiest low tech way to capture
Twitter data is using a browser

192
00:13:20,910 --> 00:13:25,350
extension. browser extensions
allow you to construct a Twitter

193
00:13:25,350 --> 00:13:29,280
search and then pull the data
off the results page into a data

194
00:13:29,280 --> 00:13:29,760
set.

195
00:13:32,480 --> 00:13:37,280
For example, data miner and
QSR’s N-capture are two that I

196
00:13:37,280 --> 00:13:40,190
have used with some success and
can both be added to Google

197
00:13:40,190 --> 00:13:45,620
Chrome. These will also scrape
data from Facebook, YouTube and

198
00:13:45,620 --> 00:13:51,530
general web pages. You have to
go through the Twitter web

199
00:13:51,530 --> 00:13:56,870
interface web search to locate
the tweets first. And the number

200
00:13:56,870 --> 00:14:00,110
of and selection of tweets
returned by the Twitter search

201
00:14:00,140 --> 00:14:02,720
through the web interface is
determined by the Twitter

202
00:14:02,720 --> 00:14:06,680
algorithms. And you may not get
all relevant tweets in your time

203
00:14:06,680 --> 00:14:09,470
period. So you're giving up some
control over your sample.

204
00:14:10,050 --> 00:14:10,620
It's

205
00:14:10,950 --> 00:14:14,490
kind of hard to know exactly
what you're getting or why

206
00:14:14,490 --> 00:14:17,910
they're giving you what they're
giving you. Also note that the

207
00:14:17,910 --> 00:14:22,380
extension in capture is very
easy to use, but only works with

208
00:14:22,380 --> 00:14:26,820
the qualitative software package
in vivo. It creates some CV x

209
00:14:26,820 --> 00:14:28,590
files that can't be used
elsewhere.

210
00:14:34,470 --> 00:14:39,150
application program interfaces
API's are protocols that manage

211
00:14:39,150 --> 00:14:46,050
and transmit data from website
to your software. To use the

212
00:14:46,050 --> 00:14:50,460
Twitter API, you need to have a
Twitter account. Sign up as a

213
00:14:50,460 --> 00:14:54,450
developer an access fee to using
a programming language such as R

214
00:14:54,450 --> 00:15:01,950
or Python. You don't actually
need To program in order to use

215
00:15:01,950 --> 00:15:05,550
a programming language to gain
access to Twitter, start by

216
00:15:05,550 --> 00:15:09,450
choosing an application track.
Now the steps I'm going to be

217
00:15:09,450 --> 00:15:13,740
showing you are relatively new.
If you've done this before,

218
00:15:13,770 --> 00:15:18,450
there may be some new steps to
follow. When I reviewed Twitter

219
00:15:18,450 --> 00:15:21,750
data access a year or so ago,
the steps were different and

220
00:15:21,900 --> 00:15:27,300
they may change again. So start
by choosing an application

221
00:15:27,300 --> 00:15:30,420
track. The three main
application tracks on the

222
00:15:30,420 --> 00:15:33,900
Twitter site are professional
hobbyist, an academic

223
00:15:34,830 --> 00:15:38,880
professional allows you access
to everything but you need to

224
00:15:38,880 --> 00:15:45,060
pay for it and it is not cheap.
academics and hobbyists not

225
00:15:45,060 --> 00:15:48,000
counting academics who follow
the academic research track,

226
00:15:48,570 --> 00:15:51,270
both gets forwarded to the
standard application, which

227
00:15:51,270 --> 00:15:55,770
allows limited access to recent
Twitter data about the last

228
00:15:55,770 --> 00:16:01,950
seven to nine days of tweets. To
follow up for the free full

229
00:16:01,950 --> 00:16:06,120
archive access academic research
track, you need to have a

230
00:16:06,120 --> 00:16:09,300
specific clearly defined
research project which we will

231
00:16:09,300 --> 00:16:13,800
just describe as part of the
application process. You also

232
00:16:13,800 --> 00:16:16,470
need to be employed as a
researcher at an academic

233
00:16:16,470 --> 00:16:19,260
institution or a grad student
working on a thesis or

234
00:16:19,260 --> 00:16:26,520
dissertation. After you complete
the application, you'll be

235
00:16:26,520 --> 00:16:30,300
forwarded to the developer
developer portal. Right now you

236
00:16:30,300 --> 00:16:35,040
can create an app to use the old
Twitter 1.2 API, or a project to

237
00:16:35,040 --> 00:16:42,240
use the Early Access version of
2.0 api 2.0 is still evolving,

238
00:16:42,240 --> 00:16:49,800
so I used 1.2 and created an
app. Your app will generate a

239
00:16:49,800 --> 00:16:52,830
set of keys and tokens that you
will need at the step where you

240
00:16:52,860 --> 00:16:57,330
set up your data access script.
It won't save them, you can

241
00:16:57,330 --> 00:17:01,290
always regenerate a new set if
you lose them. So you generate

242
00:17:01,290 --> 00:17:05,400
your token and key keys and copy
them somewhere where you won't

243
00:17:05,400 --> 00:17:10,380
lose them. I've talked a bit
about using API's to get your

244
00:17:10,380 --> 00:17:15,390
data. To do that you need the
right software. A free widely

245
00:17:15,390 --> 00:17:18,660
used tool is r which is a free
statistical programming

246
00:17:18,660 --> 00:17:21,960
language. It's used for
procedures like manipulating

247
00:17:22,020 --> 00:17:26,160
editing and visualizing data,
computing statistics and

248
00:17:26,160 --> 00:17:30,480
accessing data through API's.
And like visual most proven

249
00:17:30,480 --> 00:17:34,500
programs you interact with are
by typing commands at a prompt

250
00:17:34,500 --> 00:17:37,470
or by writing programs, which
are groups of commands that can

251
00:17:37,470 --> 00:17:42,540
be run in sequence. However, you
don't need to know how to

252
00:17:42,540 --> 00:17:46,710
program to accomplish tasks with
our many users have written

253
00:17:46,710 --> 00:17:50,490
helpful recipes that you can
simply follow step by step to

254
00:17:50,760 --> 00:17:54,330
accomplish various tasks like
getting Twitter or YouTube data,

255
00:17:54,660 --> 00:17:58,860
or running a trend analysis.
It's linked to a couple of these

256
00:17:58,860 --> 00:18:02,760
recipes in this presentation at
the end, and there are many

257
00:18:02,760 --> 00:18:05,760
others feel free to reach out to
me if you decide to take the

258
00:18:05,760 --> 00:18:12,150
plunge into using our together
analyze data. This is just a

259
00:18:12,150 --> 00:18:15,180
screenshot to show you what the
basic our interface looks like

260
00:18:15,210 --> 00:18:18,450
see have some idea what you're
getting into. There atoms

261
00:18:18,450 --> 00:18:21,480
available such as our studio
that can make are a little nicer

262
00:18:21,480 --> 00:18:25,440
to use if you expect to be doing
a lot of work with it. But

263
00:18:25,500 --> 00:18:28,830
basically using are is a matter
of setting a prompt and typing

264
00:18:28,830 --> 00:18:33,180
in commands, or writing scripts
that can system lists of

265
00:18:33,180 --> 00:18:39,810
commands. Here's the basic our
recipe or recipe or script

266
00:18:39,840 --> 00:18:43,500
showing how to use the artweek
package to access data which is

267
00:18:43,500 --> 00:18:49,080
the method I recommend. RP is
one of many add on modules that

268
00:18:49,080 --> 00:18:53,520
are available for use to our
with our that make it easier to

269
00:18:54,210 --> 00:18:59,790
do various different tasks. You
need to start by installing it

270
00:18:59,790 --> 00:19:02,940
using the install packages
command and then load the our

271
00:19:02,940 --> 00:19:07,470
library. If you decide to use
our tweet don't start by looking

272
00:19:07,470 --> 00:19:10,350
at the our tweet official
documentation, which has a list

273
00:19:10,350 --> 00:19:14,280
of several dozen functions in
alphabetical order is just not

274
00:19:14,280 --> 00:19:19,170
very user friendly. One of the
developers, Michael Kyani has

275
00:19:19,170 --> 00:19:22,560
some very good getting started.
instructions that I'm linking

276
00:19:22,560 --> 00:19:30,060
here. Okay, that's it for
Twitter. Next I'm going to be

277
00:19:30,060 --> 00:19:31,080
talking about YouTube.

278
00:19:33,119 --> 00:19:33,749
Now,

279
00:19:35,069 --> 00:19:39,449
video data is pretty challenging
to work with analysis generally

280
00:19:39,479 --> 00:19:44,609
needs to be done by hand,
watching videos slowly and

281
00:19:44,609 --> 00:19:49,529
taking notes. However, like
other platforms, YouTube also

282
00:19:49,529 --> 00:19:53,549
has plenty of textual data and
metadata. And this textual data

283
00:19:53,549 --> 00:19:58,259
can be analyzed much like other
forms of text data. comments and

284
00:19:58,259 --> 00:20:02,189
videos provide a rich corpus
Texture analysis. These can be

285
00:20:02,489 --> 00:20:05,279
analyzed in conjunction with
things like viewer, viewer and

286
00:20:05,279 --> 00:20:13,559
channel statistics. Google owns
YouTube. So the first step is to

287
00:20:13,559 --> 00:20:18,359
set up access to the Google API.
And you need to be logged in

288
00:20:18,359 --> 00:20:26,009
with a Google or Gmail account
to do this. So first, to enable

289
00:20:26,099 --> 00:20:30,119
API's on the Google API console.
I'll be showing you how to

290
00:20:30,119 --> 00:20:37,799
access these API's using the our
tuber package. actually doing

291
00:20:37,799 --> 00:20:42,359
this is even more complex than
working with Twitter because the

292
00:20:42,359 --> 00:20:46,169
Google Cloud platform provides
access to a number of API's and

293
00:20:46,169 --> 00:20:51,299
tools aside from YouTube. So you
basically need to search for the

294
00:20:51,299 --> 00:20:56,099
YouTube API, and it will
eventually come up. I'll note

295
00:20:56,099 --> 00:20:59,129
that many of the instructions I
looked at recommended also

296
00:20:59,129 --> 00:21:02,219
enabling something called the
Freebase API,

297
00:21:03,569 --> 00:21:04,289
whoever

298
00:21:04,360 --> 00:21:07,810
this has been deprecated and no
longer comes up in a search.

299
00:21:09,400 --> 00:21:12,940
From what I could find out,
extremely basic things seem to

300
00:21:12,940 --> 00:21:16,750
work without it. But I'm not
sure exactly what it did, or if

301
00:21:16,750 --> 00:21:25,090
not having it will break older
example scripts. After you

302
00:21:25,120 --> 00:21:30,250
enable the YouTube API is you
need to create an OAuth client

303
00:21:30,250 --> 00:21:34,780
ID to use in your scripts.
Here's another place where I had

304
00:21:35,260 --> 00:21:39,010
trouble with the online
instructions I found. They told

305
00:21:39,010 --> 00:21:42,280
me to select the application
type of other to create the

306
00:21:42,280 --> 00:21:46,810
correct correct type of client
ID. Whoever other no longer

307
00:21:46,810 --> 00:21:51,070
appears as an option, some web
searching later and I found that

308
00:21:51,070 --> 00:21:54,700
the desktop app is the new other
so this is the one that you need

309
00:21:54,700 --> 00:22:04,060
to select. The most widely used
our pack ship package for

310
00:22:04,060 --> 00:22:08,560
accessing YouTube data is tuber.
It works much like our fleet

311
00:22:08,620 --> 00:22:11,650
with the user set user setting
up a connection using the

312
00:22:11,650 --> 00:22:15,550
credentials they generated. And
then using simple commands like

313
00:22:15,610 --> 00:22:18,910
get all comments to get all the
comments on a specific video

314
00:22:19,990 --> 00:22:24,190
using the video ID which can be
found in the URL to a specific

315
00:22:24,190 --> 00:22:32,920
video. The last two social media
programs, platforms we'll be

316
00:22:32,920 --> 00:22:36,160
discussing are Facebook and
Reddit. While there are many

317
00:22:36,160 --> 00:22:38,920
others have tried to cover a
selection of the most widely

318
00:22:38,920 --> 00:22:43,540
used platforms that actually
allow some sort of data mining,

319
00:22:43,750 --> 00:22:46,750
and also cover a range of
different types of social media.

320
00:22:49,630 --> 00:22:53,020
Reddit is a social media news
and discussion site with some

321
00:22:53,020 --> 00:22:56,680
forums called subreddits. on
everything from politics to

322
00:22:56,680 --> 00:23:01,690
programming to cat pictures.
Some Reddit forums are rather

323
00:23:01,690 --> 00:23:05,980
controversial, outright
offensive, or simply illegal.

324
00:23:06,790 --> 00:23:09,970
For example, the subreddit that
has tips on shoplifting without

325
00:23:09,970 --> 00:23:15,130
getting caught. And Reddit has
been inconsistent about banning

326
00:23:15,130 --> 00:23:19,690
these they occasionally decide
to purge one or more of them and

327
00:23:19,690 --> 00:23:24,850
let others pop up. As a site
that relies on users for content

328
00:23:24,880 --> 00:23:28,300
interaction and moderation, some
interesting data can be pulled

329
00:23:28,300 --> 00:23:32,140
from Reddit that can show real
time trends of popular topics.

330
00:23:32,740 --> 00:23:37,000
Users upvote and downvote vote
each other's comments, with

331
00:23:37,000 --> 00:23:40,330
uploaded comments rising higher
in the threads. so popular

332
00:23:40,330 --> 00:23:43,570
comment is much more likely to
be seen by another user, and

333
00:23:43,570 --> 00:23:48,580
perhaps then further uploaded or
commented on. This makes for a

334
00:23:48,580 --> 00:23:52,270
complex network of users and
discussions that also provides

335
00:23:52,270 --> 00:23:59,290
insight into trending topics and
into how popular particularly or

336
00:23:59,290 --> 00:24:03,640
agreed with a particular
opinions on the mark. All these

337
00:24:03,640 --> 00:24:07,360
factors make Reddit a rich
source of material for analysis

338
00:24:07,360 --> 00:24:13,180
of real time trends. In popular
topics. I suggest accessing

339
00:24:13,180 --> 00:24:19,030
Reddit  using the R package Re
ditExtractoR to use this you d

340
00:24:19,030 --> 00:24:22,270
n't have to set up an authe
tication or create any kind 

341
00:24:22,300 --> 00:24:25,930
f account to access the Reddi
 API. Yay.

342
00:24:31,450 --> 00:24:34,900
Now, unlike the rest of the
platforms, I've covered Facebook

343
00:24:34,900 --> 00:24:39,190
did not allow direct access to
user data by researchers through

344
00:24:39,190 --> 00:24:42,640
any sort of API. But I still
wanted to mention that because

345
00:24:42,640 --> 00:24:46,360
it does provide access to some
data. And as a very influential

346
00:24:46,360 --> 00:24:50,470
platform. The Terms of Service
state you may not access or

347
00:24:50,470 --> 00:24:54,130
collect data from our products
using automated means without

348
00:24:54,130 --> 00:24:57,850
our prior permission or attempt
to access data that you do not

349
00:24:57,850 --> 00:25:03,310
have permission to access
However, they do have some data

350
00:25:03,310 --> 00:25:05,770
that is made available
specifically for academic

351
00:25:05,770 --> 00:25:08,800
researchers, including the
Facebook ad library,

352
00:25:09,340 --> 00:25:13,870
international future business
survey, and some datasets shared

353
00:25:13,870 --> 00:25:18,610
on the humanitarian data, our
humanitarian data exchange,

354
00:25:19,210 --> 00:25:23,260
including some COVID-19 symptom
surveys. There's also

355
00:25:23,260 --> 00:25:26,950
crowdtangle, which is a Facebook
developed tool for analyzing

356
00:25:26,950 --> 00:25:31,540
public pages, public groups,
verified profiles and public

357
00:25:31,540 --> 00:25:37,810
Instagram accounts. This data is
accessed in a variety of ways

358
00:25:37,810 --> 00:25:41,800
humanitarian data can be
downloaded directly to excel.

359
00:25:42,610 --> 00:25:45,970
The business survey requires
application and approved

360
00:25:45,970 --> 00:25:49,840
product, and the crown tip
crowdtangle data is accessed

361
00:25:49,840 --> 00:25:57,700
through a Facebook developer
interface. According to

362
00:25:57,700 --> 00:26:00,460
Facebook, the app library
provides advertising

363
00:26:00,460 --> 00:26:03,460
transparency by offering a
comprehensive searchable

364
00:26:03,460 --> 00:26:06,640
collection of all ads currently
running from across Facebook

365
00:26:06,640 --> 00:26:11,200
products, including Instagram.
Facebook provides both a

366
00:26:11,200 --> 00:26:14,260
searchable and browsable
interface to the app library and

367
00:26:14,290 --> 00:26:18,640
API access. To get API access
however, you need to sign up for

368
00:26:18,640 --> 00:26:22,030
a Facebook developer account. So
I didn't didn't know how to get

369
00:26:22,030 --> 00:26:30,790
exactly how that one works any
further. Now what you've got

370
00:26:30,790 --> 00:26:36,010
your social media data, what are
you going to do with it? I

371
00:26:36,010 --> 00:26:39,280
wanted to introduce you to some
of the very popular types of

372
00:26:39,280 --> 00:26:43,090
research that social media data
like Twitter's used for, this

373
00:26:43,090 --> 00:26:46,300
doesn't cover everything. But
these are some of the most

374
00:26:46,300 --> 00:26:49,120
popular types of research that
you're likely to come across in

375
00:26:49,120 --> 00:26:55,000
the literature. network analysis
is finding out who follows who

376
00:26:55,120 --> 00:27:00,100
who talks about what and tracing
how ideas spread. through social

377
00:27:00,100 --> 00:27:03,100
media networks, it's widely used
in marketing as well as in

378
00:27:03,100 --> 00:27:07,420
political science. trend
analysis looks at the popularity

379
00:27:07,420 --> 00:27:10,780
of concepts and hashtags over
time and can be used to relate

380
00:27:10,780 --> 00:27:14,830
opinions to events, or see what
events resonate most with the

381
00:27:14,830 --> 00:27:19,540
general public. trend analysis
is a fairly simple type of

382
00:27:19,540 --> 00:27:25,000
analysis. It can be useful to
supplement a larger project, but

383
00:27:25,030 --> 00:27:27,460
there's only a certain amount of
insight you can gain from it.

384
00:27:29,200 --> 00:27:32,950
sentiment analysis is another
fairly simple analysis that's

385
00:27:32,950 --> 00:27:38,680
used to automatically classify
tweets or other short

386
00:27:38,680 --> 00:27:44,080
expressions on social media.
associated with particular topic

387
00:27:44,080 --> 00:27:48,310
as positive, negative or
neutral. And lastly, content

388
00:27:48,310 --> 00:27:53,380
analysis is often used by
qualitative researchers to come

389
00:27:53,410 --> 00:27:58,090
summarize complex data and find
themes or common concepts that

390
00:27:58,090 --> 00:28:03,640
dominate a particular discourse.
content analysis is a way of

391
00:28:03,640 --> 00:28:06,880
teasing out the commonalities
and a massive disparate text so

392
00:28:06,880 --> 00:28:10,540
you can make inferences about
the messages in the texts, and

393
00:28:10,540 --> 00:28:15,010
is suitable for doing very
complex types of research.

394
00:28:17,620 --> 00:28:20,260
Here's an example of a trend
analysis that was done using

395
00:28:20,260 --> 00:28:23,770
Twitter data. The researcher
looked for posts mentioning

396
00:28:23,770 --> 00:28:27,970
various race or ethnicity
related keywords to see which

397
00:28:27,970 --> 00:28:31,960
one of a set of events caused
the most discourse around the

398
00:28:31,960 --> 00:28:39,910
topic. content analysis is
frequently used with social

399
00:28:39,910 --> 00:28:44,290
media data. It's widely used in
qualitative research anytime the

400
00:28:44,290 --> 00:28:47,530
researcher needs to make sense
of a large amount of textual

401
00:28:47,800 --> 00:28:52,840
voice or video data. content
analysis is basically a

402
00:28:52,840 --> 00:28:56,830
formalized way of summarizing a
massive text data to make it

403
00:28:56,830 --> 00:29:00,880
easier to discover meaning
chunks of text phrases or

404
00:29:00,970 --> 00:29:04,360
sentences and interview data,
for example, or individual

405
00:29:04,360 --> 00:29:08,680
tweets are tagged with
descriptors that classify them

406
00:29:08,680 --> 00:29:12,040
according to some schema
classification. Classification

407
00:29:12,040 --> 00:29:15,070
schemes can be a priori, which
means you come up with a list of

408
00:29:15,070 --> 00:29:19,360
terms according to some theory
before analysis, or inductive,

409
00:29:19,870 --> 00:29:22,660
which means it's made up from
terms discovered by looking

410
00:29:22,660 --> 00:29:24,160
through a sub sample of the
data.

411
00:29:28,830 --> 00:29:32,100
The research focus in content
analysis can be more

412
00:29:32,100 --> 00:29:35,970
quantitative, focused on
counting co authored particular

413
00:29:35,970 --> 00:29:40,260
themes occur in the data, or it
can be more qualitative, focused

414
00:29:40,530 --> 00:29:45,000
on identifying and describing
key themes. This is also called

415
00:29:45,030 --> 00:29:50,220
thematic analysis. Another
division is between relational

416
00:29:50,220 --> 00:29:53,490
call and text analysis looking
at which concepts tend to occur

417
00:29:53,490 --> 00:29:57,540
together in touch well, which is
just coding for simple concepts.

418
00:29:58,440 --> 00:30:02,100
Social media like tweets and
Comments may be more suited for

419
00:30:02,100 --> 00:30:05,730
conceptual analysis being
shortened, often focused on

420
00:30:06,420 --> 00:30:11,220
indicating a single key concept.
I know I'm not really giving you

421
00:30:11,220 --> 00:30:14,550
enough detail to go out there
and start conducting a content

422
00:30:14,550 --> 00:30:18,420
analysis of your own. But I did
just want to introduce these

423
00:30:18,420 --> 00:30:22,200
basic concepts to give you an
idea of what types of keywords

424
00:30:22,200 --> 00:30:24,810
to search for when you want to
figure out exactly how point

425
00:30:24,810 --> 00:30:30,540
analyze your social media data.
Here's one popular way you might

426
00:30:31,170 --> 00:30:34,740
possibly you might conduct a
content analysis of some Twitter

427
00:30:34,740 --> 00:30:40,320
data. Imagine our research
question is how do Twitter users

428
00:30:40,320 --> 00:30:45,240
use mood music to deal with
COVID-19 anxiety? The first step

429
00:30:45,270 --> 00:30:49,860
in getting the data is to choose
a hashtag phrase or set of terms

430
00:30:49,890 --> 00:30:52,050
relating to your research
question that you want to

431
00:30:52,050 --> 00:30:56,430
investigate. A search trip
string might include COVID-19,

432
00:30:56,460 --> 00:31:01,170
or COVID, or Corona, music or
playlist, anxiety or fear or

433
00:31:01,170 --> 00:31:08,700
worry. Next to use these terms
to retrieve a set of data using

434
00:31:08,730 --> 00:31:13,290
our tweet or some other package.
And you would have to decide on

435
00:31:13,290 --> 00:31:16,170
a sample size for example, you
might decide to look at the

436
00:31:16,170 --> 00:31:21,690
first 500 tweets retrieved by
your search. You'd extract the

437
00:31:21,690 --> 00:31:26,250
data and import the sample into
Excel or in vivo or some other

438
00:31:26,250 --> 00:31:30,180
software package. And then the
most time consuming part,

439
00:31:30,420 --> 00:31:34,380
reading through your data to see
what groupings emerge and

440
00:31:34,380 --> 00:31:37,410
tagging each tweet with a
descriptive word or phrase

441
00:31:37,410 --> 00:31:41,010
summarizing its theme. Or in
some cases, you might want to

442
00:31:41,010 --> 00:31:44,910
tag it as a relevant or an
interpretable. Finally, you

443
00:31:44,910 --> 00:31:48,810
would group them condense your
descriptors into broad themes,

444
00:31:48,840 --> 00:31:51,540
use these themes to summarize
what you've learned about your

445
00:31:51,540 --> 00:31:56,340
research question. Here's an
example of a fairly simple

446
00:31:56,340 --> 00:32:00,000
content analysis that was done
to classify tweets on the H1N1

447
00:32:00,000 --> 00:32:04,140
Epidemic  by type. This
particular analysis questions

448
00:32:04,200 --> 00:32:08,340
the tweets into groups such as
resource, personal experience,

449
00:32:08,370 --> 00:32:12,630
and jokes. analyses with a
different focus might have

450
00:32:12,720 --> 00:32:16,080
classified them based on
emotional content positivity, or

451
00:32:16,080 --> 00:32:20,220
negativity, tivity, or some
other schema. There are always

452
00:32:20,220 --> 00:32:24,990
many different ways to classify
any given set of data, depending

453
00:32:24,990 --> 00:32:31,920
on what your research question
is. Now, that's all I've

454
00:32:31,920 --> 00:32:34,590
prepared for you today. But if
you're interested in the topics

455
00:32:34,590 --> 00:32:37,350
I've discussed, there are plenty
of online sources where you can

456
00:32:37,350 --> 00:32:41,280
find more information. I've
compiled a selection of my

457
00:32:41,280 --> 00:32:46,890
favorite links is starting
points. And, again, I'm Kristy

458
00:32:46,890 --> 00:32:50,460
Thompson, research data
librarian and if you have any

459
00:32:50,460 --> 00:32:54,060
questions or want support with
any of the topics I've talked

460
00:32:54,060 --> 00:32:59,160
about, you can contact me
directly at  kthom67 at uwo dot

461
00:32:59,160 --> 00:33:03,360
ca or reach out to the research
and scholarly communication team

462
00:33:03,420 --> 00:33:07,830
at RSC lib@uwo.ca. Thank you for
listening