Swift Tracker Methodology

14th April 2015 Uncategorized

There are no clever algorithms in our research on Taylor Swift, just search terms and manual data entry into an unattractive spreadsheet. We decided not to weight any of our data. Aggregations are made purely on the numbers of direct returns from our searches.

Nexis returns are likely to have a margin of error. Below 20 returns we weeded out references to eg. Hozier Street, or Clementine Hozier, but above that number we let the errors stand. For Hozier in particular that means there is a probable margin of error of 5/6 returns on mentions above 20 since this was the average number of references to these other irrelevant Hoziers in <20 returns. There may be some similar errors for Sam Smith and Haim, though to a lesser degree.

Search terms on Nexis were as follows:

  1. Famous friends: Ed Sheeran, Sam Smith singer, Hozier, Lorde singer, Haim the band, Cara Delevingne, Lena Dunham and Harry Styles
    1. Anywhere in the text
    2. Quarterly 2010-2015 ending 31 March 2015
    3. All English Language News
    4. Group duplicates – moderate
  2. Issues: feminism, feminist, misogyny, mansplaining, transgender, LGBT, hipster
    1. In the headline
    2. Quarterly 2010-2015 ending 31 March 2015
    3. All English Language News
    4. Group duplicates – moderate
  3. Taylor Swift
    1. In the headline
    2. Quarterly 2010-2015 ending 31 March 2015
    3. All English Language News
    4. Group duplicates – moderate
  4. Taylor Swift AND each separate search term
    1. Anywhere in the text
    2. Quarterly 2010-2015 ending 31 March 2015
    3. All English Language News
    4. Group duplicates – moderate

A Nexis source list is available here. Nexis mainly tracks established news sources like broadsheet newspapers but does also include data from major news websites.

 

Historic Twitter searches are rather hard to do without access to the ‘hosepipe’, which we were not able to gain. We used a workaround through Google Search that seemed to give us returns in the right proportions. We aren’t absolutely certain exactly what these numbers are tracking but they’re definitely mentions of our search terms in the right years.

The search criteria looked like this:

“X”/”@X” site:twitter.com daterange:YYYYYYY-YYYYYYY

This advanced search includes several operators: X represents the search term (see below), Y represents the date range indicated by a Julian calendar reference (created using this website), and the website to search within is also included (twitter.com).

The search terms used in Google search for Twitter were as follows:

  1. Famous friends: @edsheeran, @samsmithWorld, @hozier, @lordemusic , @haimtheband, @caradelevingne, @lenadunham, @harry_styles
  1. Issues: feminism, feminist, misogyny, mansplaining, transgender, LGBT, hipster

They do not differ from the Nexis search terms other than that we used Twitter handles rather than names for the ‘famous friends’ category. This method tends to yield more results when searching in Twitter, however, a more comprehensive search might aggregate mentions of names as well as handles (‘indirect’ tweets), accounting for the fact that some Tweets may feature both.

 

When searching for Swift’s direct interactions with our search terms we used the Twitter advanced search tool and Snap Bird. With Twitter’s advanced search we visually extracted Swift’s mentions of the search terms below. With Snap Bird we were able to search in Swift’s favourited Tweets and also visually count the number of times Swift favourited a tweet that either mentioned one of our search terms or was originally Tweeted from one of the relevant handles.

The search terms used for Swift’s Twitter interactions were as follows:

  1. Famous friends: @edsheeran, @samsmithWorld, @hozier, @lordemusic , @haimtheband, @caradelevingne, @lenadunham, @harry_styles
  1. Issues: feminism, feminist, misogyny, mansplaining, transgender, LGBT, gay icon*, hipster

*Due to the visual extraction of data from Swift’s tweets we were able to include in our data Swift’s favourited Tweet which included the phrase ‘gay icon’ which we felt was enough in the spirit of the ‘LGBT’ search term.

For completeness, and in case anyone is interested, a full range of Taylor Swift charts is available. These are published here under a creative commons license, so do please credit our hard work on the data sources, if you would like to use them, with a link to osca.co

 

– Genevieve Maitalnd Hudson and Joanna Weir