Thursday, 18 June 2015

Investigating Language on Twitter

Introduction

we set out a hypothesis that men are more dominant as they are more straight to the point whereas women tend to waffle which links to the deficit which is a power theory. we planned on using one man and one women who were both in the public eye, these were Ed Milliband- an MP and former leader of the labour party, and Amy Childs - a reality TV star

Methodology

our methodology was two pick every third tweet from each of these people. We chose to do it this way to prove that there are no anomalies. we then proceeded to count the number of each various feature of each tweet to try and prove our hypothesis.

Analysis

The results are as follows:

  • Emoji's men used 0 and women used 12
  • Emotive language men used 13 and women used 7
  • Men's sentence lengths ranged for men at 16-27 with an average of 20.1. Women on the other hand had a range of 3-17 with an overall average of 9.2
  • Amount of hashtags used by men is 1 and women was 10
we have ultimately found out that our findings do not support our hypothesis. this could primarily be because of the choice of the people in which we chose to analyse . Ed Milliband is a politician and Amy Childs is a reality TV star. This shows that these two are from very different backgrounds and therefore it is unlikely that they can be compared in this way and be a general stereo-type   of each of their gender groups.

Conclusion

Reflecting on this study I think its clear that we may have chosen the wrong subjects for this investigation. If this this investigation were to be done again we could choose people from different similar backgrounds because if we did this the two people would be easier to compare and hopefully give us more conclusive evidence for our hypothesis.

1 comment:

  1. Some good exploration of the data you chose and some good evaluation. I think you need to use terms like are they 'representative' of their gender and is the data 'reliable' and 'generalisable'. Random selection won't avoid anomalies but it will avoid biased selection. A larger data pool minimises the effect of anomalies on quantified data and allows you to spot them more easily. Be careful with subjective terms like 'emotive language' and find ways to objectively measure this, evaluating the success of this approach as you go by looking at examples in context, PEE.

    ReplyDelete