We’re having quite a busy fall here at Etuma, so none of us have had as much time as we had wanted to do in-depth analysis on the presidential candidates’ Facebook posts nor to write the Pulitzer-worthy posts we had hoped for. But there is a lot of interesting data there, and we want to make it available to you with or without our own brand of investigative journalism. So, today I decide to just throw up some of the analysis straight from our tools and let you draw your own conclusions. I picked the topic ‘Truthfulness’, because it has been quite active from the beginning and I wanted to see what kind of comments were behind it. And now, without further ado…
Weekly volumes for the topic ‘Truthfulness’ by week number
The topic peaked a couple of weeks ago, but keep in mind that the overall comment volume also peaked a couple of weeks ago.
Keyword tag cloud for topic ‘Truthfulness’
Cumulative volume and sentiment by recipient
Remember that +1 would be the result of all positive comments and -1 all negative. Also don’t forget that anyone is free to post a comment about either candidate on both Facebook walls. So, it is difficult to know if we can draw some conclusions from the fact that one candidate’s wall has a slightly more negative sentiment than the other. But it is clear that this is a hot topic on both sides of the electorate.
We’ll be back soon with more analysis and interesting findings, but in the meantime, let us know what you would like to see. Is there any specific topic you’ve wondered about? Post your suggestions in the comments section or tweet @baramitter.
I’d be interested in seeing how your data correlates to polls. One way of looking at polls is that it’s a kind of “actual” or “truth” about the public opinion at any given time. That presents a fairly unique opportunity to compare the results arrived at through social media sentiment mining and refinement with the “actuals” that are revealed through polls. Further, as presidential election polls are conducted at quite a high frequency, it should be possible to not just compare your data to the “truth” as a single bulk comparison, but instead at multiple points in time. Given the right analysis and refinement, this should yield a high level of correlation that persists over time.
Thanks for the comment Peppe. We have also been discussing internally that we need to concentrate on trend analysis more than the bulk comparisons we have been doing so far. Comparing the data to polls is a good idea as well. Our originally idea was to choose a few news headlines each week to which we would compare our data, but perhaps polls would yield a better base for comparison to our findings.