For my final year project, I created a machine-learning based pipeline to analyse sentiment on Twitter in real-time regarding any particular topic. As an initial test case, I had it monitor the 8th amendment debate in Ireland for a few months leading up to the vote. The results of the analysis were displayed live on this site. Now that the debate is over and the pipeline is no longer running, this site now displays a cached version of the results it generated.
Ireland is set to hold a referendum on whether or not to repeal the 8th amendment of the Irish constitution. This would remove the effective ban on abortion in Ireland. The 2 sides of the debate are those who want to repeal the 8th amendment (ie 'Pro-Choice') and those who want to save the 8th amendment (ie 'Pro-Life')
Mood is a generalisation of the main emotion appearing in a tweet. Positive mood (with 1 being a totally positive score) includes happiness, hope, kindness... Negative mood (with -1 being a totally negative score) includes sadness, despair, anger...
Given abortion is such a divisive topic, it's likely to cause strong outbursts of mood from both ends of the spectrum, and from both sides of the debate.
An example of a tweet with a positive mood could be "Loving this article that appeared in the Dungarvan Leader! #savelives #savethe8th"
A tweet with negative mood would be more like "Anyone who is pro-abortion needs to go back to primary school and study basic science. You might then be able to put your brain cells together and realize that abortion is the murder of an innocent child.#ProLife #savethe8th"
The wordclouds display the words currently most used by each side, relative to the other. This allows us to see what topics are being focused on by each side, and also the different type of language used. As opposed to the mood graph which gives an overview of the entire debate, the wordclouds focus on the current state of the debate and the trending topics.