Social media is now the place where people are gathering en masse to discuss the news with their friends, neighbors and complete strangers. This change in news consumers’ behavior is proving to be a challenge for local news, but it is also an opportunity. Users and system generated data from social media can also be a boon for content creators. This presentation will feature a case study showing how publishers can use social media analytics to gain insights into their audience and how to use this information to foster a stronger sense of community around their brand of journalism. The case study will focus on how to use Netlytic, a cloud-based social media analytics tool, to mine the public Facebook interactions of the readers of BlogTO, a regional, Canadian-based media outlet, to find out what their readers are interested in and what engages them.
Online Social Shopping Motivation: A Preliminary Study
Predicting what gets ‘Likes’ on Facebook: case study of BlogTO
1. Predicting what gets ‘Likes’ on Facebook:
A case study of BlogTO
Philip Mai (@phmai)
Director, Business & Communications
Social Media Lab
Ryerson University
Priya Kumar (@link_priya)
Postdoctoral Fellow
Social Media Lab
Ryerson University
2. About BlogTO
“Toronto's source for local
news and culture,
restaurant reviews, event
listings and the best of
the city.”
Facebook Page
(est. in 2004)
~300K Followers
@SMLabTO 2
4. Research Questions
• What kinds of BlogTO posts get more likes on Facebook?
1. Are there certain types of BlogTO content (videos, photos, links,
events, status updates) that are more engaging for readers?
2. Are there linguistic cues that can predict the ‘likeability’ of
BlogTO’s posted content on Facebook?
@SMLabTO 4
5. Methodological Toolkit
Data
Collection:
Netlytic
• Social media data collector
• Text and network analyzer
Text
Analysis:
LIWC
• Automated text analysis (LIWC –
Linguistic Inquiry & Word Count )
Statistical
Analysis:
SPSS
• Statistical software
@SMLabTO 5
6. - a social media analytics platform
designed for researchers to
collect, analyze and visualize
publicly available data from…
• Twitter, Youtube, Instagram,
Facebook, blogs, etc…
• Used by thousands of
students & scholars
Netlytic.org
@SMLabTO 6
7. Visualize & analyze social networks
Discover popular topicsCollect data from social media
Find & explore emerging
themes of discussions
@SMLabTO 7
Data Collection &
Analysis
Netlytic.org
9. Sample Facebook Data viewed in Excel
pubdate author post type like_count
4/24/2017
17:46:00
blogTO Toronto looking like a Unicorn Frappuccino - Photo by
alexandramack22
photo 6431
4/4/2017
18:31:00
blogTO Some of Toronto's favourite food vendors are now in
one place
video 5701
4/7/2017
18:31:00
blogTO Toronto has a new spot for epic ice cream treats video 5225
4/10/2017
14:16:41
blogTO Mark your calendars link 4894
4/15/2017
18:31:00
blogTO Toronto just got a secret superhero and villain themed
restaurant
video 4378
4/8/2017
8:31:00
blogTO Good morning! - Photo by zzoomed photo 4066
@SMLabTO 9
10. BlogTO
Dataset
April 3 – May 3 2017
11,785 Unique Posters
17,748 Posts +
Replies
“Who Replies To Whom” Network
(no reciprocal ties)
@SMLabTO 10
Data Collection
Netlytic.org
11. BlogTO Dataset
April 3 – May 3 2017
11,785 Unique Posters
17,748 Posts +
Replies
641 BlogTO
Posts Focus of
Text Analysis
@SMLabTO 11
12. Collected Data & Metadata as Captured by Netlytic
Sample Post
post Toronto looking like a
Unicorn Frappuccino -
Photo by alexandramack22
date 4/24/2017 17:46:00
author blogTO
type photo
like_count 6431
link https://www.facebook.com/blogt
o/posts/10154396883870009 @SMLabTO 12
13. Methodological Toolkit
Data
Collection:
Netlytic
• Social media data collector
• Text and network analyzer
Text
Analysis:
LIWC
• Automated text analysis (LIWC –
Linguistic Inquiry & Word Count )
Statistical
Analysis:
SPSS
• Statistical software
@SMLabTO 13
14. The Power of Words
A lot has been written about the power of words to drive
change, acquire customers, and persuade crowds …
What is the role of language in driving
engagement and traffic to local news sites?
@SMLabTO 14
15. Linguistic Inquiry and Word Count (LIWC)
LIWC Dictionary Contains 90 Word Categories
Such As:
• Linguistic dimensions (articles, verbs)
• Psychological constructs (affect, cognition)
• Personal concerns (work, leisure)
• Informal language (swear words) online
speech
• Punctuation (periods commas)
LIWC
Sample LIWC Categories (90 in total)
• “Risk” - words that are perceived as
threatening
• “Power” - words that signify strength or
control
• “Leisure” - words that refer to activities
not associated with work
@SMLabTO 15
16. Text Analysis with LIWC
BlogTO Facebook Posts collected by
Netlytic
LIWC Output: posts & corresponding scores
for 90 categories
post type
Toronto looking like a Unicorn
Frappuccino - Photo by
alexandramack22
photo
Some of Toronto's favourite food
vendors are now in one place
video
Toronto has a new spot for epic ice
cream treats
video
Mark your calendars link
Toronto just got a secret superhero
and villain themed restaurant
video
Good morning! - Photo by zzoomed photo
@SMLabTO 16
18. LIWC Categories
Sample Post
post Toronto looking like a Unicorn
Frappuccino - Photo by
alexandramack22
LIWC Category Count Score %
# of Words 9 100%
Function
3 33%
Prep
2 22%
Percept
2 22%
See
2 22%
Article
1 11%
Verb
1 11%
Compare
1 11%
Used in the
analysis
@SMLabTO 18
19. Methodological Toolkit
Data
Collection:
Netlytic
• Social media data collector
• Text and network analyzer
Text
Analysis:
LIWC
• Automated text analysis (LIWC –
Linguistic Inquiry & Word Count )
Statistical
Analysis:
SPSS
• Statistical software
@SMLabTO 19
20. Using SPSS to Predict BlogTO Facebook Likes
We tested post type( videos, photo, link, etc…)+ 90 LIWC dictionary
categories as possible predictors of #Likes
FB Like
Count
Post Type
Social Processes
Informal Language
Function Words
Comparisons
Discrepancy
Analytic Thinking
Affiliation
Interrogatives
Netspeak
buddy, coworker, mom …
video, photo, link, event, share
OK, ummm, blah …
pronouns, prepositions, articles …
greater, best, more than …
should, would, could …
logical and hierarchical thinking
e.g., buddy, coworker, mom, brother…
how, when, what, where, why …
thx, btw, brb…
@SMLabTO 20
21. Result: Using SPSS to Predict BlogTO Facebook Likes
10 (of 90) LIWC Dictionary Categories were found to be Statistically
Significant (p<.05)
FB Like
Count
Post Type = Video (8x more influential than the next category)
Social Processes
Informal Language
Function Words
Comparisons
Discrepancy
Analytic Thinking
Affiliation
Interrogatives
Netspeak
R2 = 0.24
@SMLabTO 21
23. Example of Facebook Post with High
‘Social Processes’ (ex: buddy, coworker, mom)
Additional sample FB posts
in this category
1. “Show mom some love”
2. “Attention parents!”
3. “For your next date night”
@SMLabTO 23
24. Example of Facebook Post with High
‘Informal Language’ (ex: OK, ummm, blah)
Additional sample FB posts
in this category
1. “Yes, yes, yes!”
2. “Try ‘em all”
3. “FYI”
@SMLabTO 24
25. Example of Facebook Post with High
‘Netspeak’ (ex: thx, btw, brb)
Additional sample FB posts
in this category
1. “Yup”
2. “Hmmm”
3. “Awww”
@SMLabTO 25
26. Implications
Be …
Engage with your audience through posts & replies (not just shares)
Other studies showed that it’ll help to build a community and not just attract followers
Conversational
Use more videos! Photos are so last year?Visual
Embrace informal language but avoid netspeak;Informal
Use posts that project to future or direct to activity
e.g., “this should be good”, “this is a must-see”
Future-
forward
@SMLaTO 26
27. Future Research
1
Analyze the content
of photos, videos
and blogs shared on
Facebook (not just
Facebook textual
posts)
2
Analyze who is
engaging with the
posts
3
Account for the
temporality and
seasonality of the
Toronto–scene
@SMLabTO 27
28. Predicting what gets ‘Likes’ on Facebook:
A case study of BlogTO
Philip Mai (@phmai)
Director, Business & Communications
Social Media Lab
Ryerson University
Priya Kumar (@link_priya)
Postdoctoral Fellow
Social Media Lab
Ryerson University