19
The Dynamics of Topical Conversations in Microblogging Victoria Lai and William Rand SocialCom 2013 September 11, 2013

Does Love Change on Twitter? The Dynamics of Topical Conversations in Microblogging

  • Upload
    vl3

  • View
    42

  • Download
    1

Embed Size (px)

DESCRIPTION

Presentation at SocialCom 2013

Citation preview

Page 1: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

The Dynamics of Topical Conversations in Microblogging

Victoria Lai and William RandSocialCom 2013September 11, 2013

Page 2: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Keyword and trend identification Background noise Importance of time, subject, geography?

2

Page 3: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Twitter’s trending topics Real-time trend identification◦ Mathioudakis and Koudas (2010)◦ Cataldi, Di Caro, and Schifanella (2010)

Other trend detection methods◦ Benhardus and Kalita (2013)

Trends in a geographic area◦ Wilkinson and Thelwall (2012)

3

Page 4: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Daily Twitter REST API queries, 47 weeks Topics◦ 11 keywords/phrases chosen for time, subject, and

geographic variety (baseball, economy, global warming, love, London riots)◦ 5 control (I, the, and, a, of)

Geography◦ Global◦ 9 cities (Boston, D.C.,

London, Sydney)

4

Page 5: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Term frequency

Inverse document frequency

TF-IDF

Page 6: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Keyword lists as target/control sets vary Time, subject, geography Tf-idf and rank correlation

𝑓𝑓 𝑇𝑇𝑡𝑡∗, 𝑠𝑠∗, 𝑔𝑔∗ 𝐶𝐶𝑡𝑡, 𝑠𝑠, 𝑔𝑔 )

6

Page 7: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

tf phrase 1 #phrase 2 #phrase 3 #C0, ∀sc, ∀g

C1, ∀sc, ∀g

C2, ∀sc, ∀g

C3, ∀sc, ∀g

idf

phrase 1 #phrase 2 #phrase 3 #

phrase 1 #phrase 2 #phrase 3 #

T0, s*, ∀g

* SC = {I, the, and, a, of }

Page 8: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Control set variation by time◦ Single week◦ Cumulative week

Control set variation by geography Target set variation by time

8

Page 9: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Topic matters, time doesn’t.t

9

Page 10: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Top keywords are very consistent, particularly

for subjective topics.

10

Page 11: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

The correlation decays as we add tweets from another week back in time.

11

Page 12: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

12

Page 13: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

For trends relative to the current background noise, we only need a few weeks of control data.

13

Page 14: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Trends are independent of local background noise and more dependent on global background noise.

14

Page 15: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Some topics are more similar to background conversations than others.

15

Page 16: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Subjective topics exhibit less vocabulary change over time.

16

Page 17: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

The results of this analysis provide suggestions for how to build a trend monitoring tool.

These recommendations help define and understand: Frequency of collection Short-term vs. long-term collections Local collections vs. global collections Keyword selection for best results

17

Page 18: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

Subject variation Other methods of comparison Blogging data

18

Page 19: Does Love Change on Twitter? The Dynamics of Topical Conversations  in Microblogging

19