Hashtag lifelines
             A digital methods summer school 2012 project

                Jill, Colleen, Diego, Johannes, Sara, Albrecht,
           Allesandro, Kalina, Tally, Esther, Noortje & Carolin
Liveliness of issue terms

  Focus on hashtag mining as
  technique for analysing variability of
  issue terms over time.




1. Are hashtags a suitable format for analysing liveliness of issue terms?
2. Does co-word provide a useful alternative to frequency measures here?


• Rather than defining what rises and falls (Downs 1974), detect what is active
and changes in association.
• Defining and applying different measures of liveliness (frequency, co-word; per
interval, per day).
The Dataset


• Twitter data for “Climate Change”.
• Period: 01.02. - 15.06
• Interval: six 2 week intervals
• Total 204795 tweets.
• Focus on hashtags.
What we did




1. Variation of key   3. Actor profiling per
hastags over time     hashtag over time




                      4. Associational
2. Mapping users      profiles per hashtag
1. Hashtags over time



• Question: What are the top hashtags per interval and how do
they vary over time based on frequency vs. co-word measures?
• Address relation between frequency & co-word measures.


METHOD: Identify top hashtags per interval (A. by
frequency and B. by co-word degree), determine
most frequent hashtags per interval for both as tube.
1. Hashtags over time
!"
                              #!!"
                                          $!!"
                                                        %!!"
                                                                    &!!"
                                                                                          '!!"
                                                                                                  (!!"
                                                                                                                    )!!"
                                                                                                                                    *!!"
                                                                                                                                            +!!"
                                                                                                                                                   #!!!"
!#,!%,#$"
!%,!%,#$"
!',!%,#$"
!),!%,#$"
!+,!%,#$"
##,!%,#$"
#%,!%,#$"
#',!%,#$"
#),!%,#$"
#+,!%,#$"
$#,!%,#$"
$%,!%,#$"




                                                                                                                                                                                                                                                                                        issue transformer)
$',!%,#$"
$),!%,#$"
$+,!%,#$"
                                                                                                                                                                                                                                                                                                                                                            Top hashtags per day.



%#,!%,#$"
!$,!&,#$"
!&,!&,#$"
!(,!&,#$"
!*,!&,#$"
                                                                                                                                                                                                                                                                                                                          • Bursts have short durations.



#!,!&,#$"
#$,!&,#$"
                                                                                                                                                                                                                                                     • The day as relevant time unit:
                                                                                                                                                                                                                         are we too focused on intervals?
                                                                                                                                                                                                                                                                                                                          • Frequency helps us understand



#&,!&,#$"
#(,!&,#$"
                                                                                                                                                                                                                                                                                        what is a hashtag (publicity device,




#*,!&,#$"
$!,!&,#$"
$$,!&,#$"
$&,!&,#$"
$(,!&,#$"
$*,!&,#$"
%!,!&,#$"
!$,!',#$"
!&,!',#$"
!(,!',#$"
!*,!',#$"
#!,!',#$"
#$,!',#$"
#&,!',#$"
#(,!',#$"
#*,!',#$"
$!,!',#$"
$$,!',#$"
$&,!',#$"
$(,!',#$"
$*,!',#$"
%!,!',#$"
!#,!(,#$"
                                                                                                                                                           !"#$"%#&'()*+,'-./01'2&%.3)4.'-56%7/.#'/89'-56%7/.#53/80#:'




!%,!(,#$"
!',!(,#$"
!),!(,#$"
!+,!(,#$"
##,!(,#$"
#%,!(,#$"
#',!(,#$"
#),!(,#$"
#+,!(,#$"
$#,!(,#$"
                                                                                           -7$"




                                                     -BCDB"


                              -HIEE@"
                                                              ->?@A?"
                                                                                                                  -14673/"




                                        -@EFGEAD@"
                                                                                                                                                                                                                                                                                                                                                                                    1. Hashtags over time




                                                                                                                             -./012./345"




            -HJDG?JF?IKL@H"
                                                                                                  -0289:32;02<"


                                                                        -6180<=01:.<9."
2. Users




Network of top connected users
2. Users

Top 60 users: human/bot proportion
                                          *+,-.%
                                            /&%
                                           0&)%

                                                                                    !"#$%
                                                                                    *+,-.%

                                                                !"#$%
                                                                 &&%
                                                                '()%




Top 60 users: bot actor types
                                                          !"#$%&#                 ,-.,/,.012#

                                     $*"#+$&#                                     -345#
                                                                                  167/,58#

                                                                   $$"#'$&#       -345##
                                                                                  9-1-6,12#
                                                                                  ,-.,/,.012#
                                       !"#$%&#
                                                           %"#$(&#                8,56#


                                                 '"#)&#'"#)&#
2. Users

Bot activity patterns
2. Users

Human activity pattern
3. Hashtag URL profiling

METHOD: Identify hashtags for URL
profiling, extract URLs.
Hashtags: #ows, #tcot
3. Hashtag URL profiling
3. Hashtag URL profiling
4. Associational profile


Hashtag lifelines by exploring
associational profiles.
Method:
• Select top-connected hashtags for
profiling.
• Create profiles for each hashtags
• Visualise as streamgraph.
4. Associational profile
4. Associational profile
4. Associational profile
4. Associational profile
Thank you.

Questions?

Hashtag lifelines

  • 1.
    Hashtag lifelines A digital methods summer school 2012 project Jill, Colleen, Diego, Johannes, Sara, Albrecht, Allesandro, Kalina, Tally, Esther, Noortje & Carolin
  • 2.
    Liveliness of issueterms Focus on hashtag mining as technique for analysing variability of issue terms over time. 1. Are hashtags a suitable format for analysing liveliness of issue terms? 2. Does co-word provide a useful alternative to frequency measures here? • Rather than defining what rises and falls (Downs 1974), detect what is active and changes in association. • Defining and applying different measures of liveliness (frequency, co-word; per interval, per day).
  • 3.
    The Dataset • Twitterdata for “Climate Change”. • Period: 01.02. - 15.06 • Interval: six 2 week intervals • Total 204795 tweets. • Focus on hashtags.
  • 4.
    What we did 1.Variation of key 3. Actor profiling per hastags over time hashtag over time 4. Associational 2. Mapping users profiles per hashtag
  • 5.
    1. Hashtags overtime • Question: What are the top hashtags per interval and how do they vary over time based on frequency vs. co-word measures? • Address relation between frequency & co-word measures. METHOD: Identify top hashtags per interval (A. by frequency and B. by co-word degree), determine most frequent hashtags per interval for both as tube.
  • 6.
  • 7.
    !" #!!" $!!" %!!" &!!" '!!" (!!" )!!" *!!" +!!" #!!!" !#,!%,#$" !%,!%,#$" !',!%,#$" !),!%,#$" !+,!%,#$" ##,!%,#$" #%,!%,#$" #',!%,#$" #),!%,#$" #+,!%,#$" $#,!%,#$" $%,!%,#$" issue transformer) $',!%,#$" $),!%,#$" $+,!%,#$" Top hashtags per day. %#,!%,#$" !$,!&,#$" !&,!&,#$" !(,!&,#$" !*,!&,#$" • Bursts have short durations. #!,!&,#$" #$,!&,#$" • The day as relevant time unit: are we too focused on intervals? • Frequency helps us understand #&,!&,#$" #(,!&,#$" what is a hashtag (publicity device, #*,!&,#$" $!,!&,#$" $$,!&,#$" $&,!&,#$" $(,!&,#$" $*,!&,#$" %!,!&,#$" !$,!',#$" !&,!',#$" !(,!',#$" !*,!',#$" #!,!',#$" #$,!',#$" #&,!',#$" #(,!',#$" #*,!',#$" $!,!',#$" $$,!',#$" $&,!',#$" $(,!',#$" $*,!',#$" %!,!',#$" !#,!(,#$" !"#$"%#&'()*+,'-./01'2&%.3)4.'-56%7/.#'/89'-56%7/.#53/80#:' !%,!(,#$" !',!(,#$" !),!(,#$" !+,!(,#$" ##,!(,#$" #%,!(,#$" #',!(,#$" #),!(,#$" #+,!(,#$" $#,!(,#$" -7$" -BCDB" -HIEE@" ->?@A?" -14673/" -@EFGEAD@" 1. Hashtags over time -./012./345" -HJDG?JF?IKL@H" -0289:32;02<" -6180<=01:.<9."
  • 8.
    2. Users Network oftop connected users
  • 9.
    2. Users Top 60users: human/bot proportion *+,-.% /&% 0&)% !"#$% *+,-.% !"#$% &&% '()% Top 60 users: bot actor types !"#$%&# ,-.,/,.012# $*"#+$&# -345# 167/,58# $$"#'$&# -345## 9-1-6,12# ,-.,/,.012# !"#$%&# %"#$(&# 8,56# '"#)&#'"#)&#
  • 10.
  • 11.
  • 12.
    3. Hashtag URLprofiling METHOD: Identify hashtags for URL profiling, extract URLs. Hashtags: #ows, #tcot
  • 13.
    3. Hashtag URLprofiling
  • 14.
    3. Hashtag URLprofiling
  • 15.
    4. Associational profile Hashtaglifelines by exploring associational profiles. Method: • Select top-connected hashtags for profiling. • Create profiles for each hashtags • Visualise as streamgraph.
  • 16.
  • 17.
  • 18.
  • 19.
  • 20.