One of the challenges of analysing textual data in the social-sciences or humanities is that data often requires a binary categorization: does it fall into this box, or that box? This can be a tough call to make, especially when working with data that needs that extra bit of context to be interpreted. And if your data set is too huge to be read through with that level of care, you might want to use a computer to answer some quick questions and give you the gist of it (by making a word cloud, for instance). One neat way to add some of that missing qualitative context is by applying a sentiment analysis to your data. What this means is, if you are already analysing your data by doing something similar to text mining, sentiment analysis can sort your text even further, using lexical signifiers of emotion, to determine if the data leans towards positive or negative emotions. This is a crowd-pleasing way of taking something that we think of as very subjective (our emotions), and boiling it down to a few abstract representations. Of course, as with all data analysis, some nuance gets lost along the way.
For example, companies might look at all their Twitter or Facebook mentions to see whether people are reacting to them in a positive or a negative light. Or voters might be interested in parsing the language of politicians. You may also notice that a version of sentiment analysis is already built in to the way that we read certain content online: Facebook asks us to ‘react’ to posts, and sites like Buzzfeed sort their content based on emotional response, rather than more traditional news categories.
Screengrab via Buzzfeed
We will be using sentiment analysis in our climate change data to explore interviewee sentiments around the concept of change.
How have you encountered or used sentiment analysis? Do you think it adds or subtracts from the way we read and analyse? Tweet us your responses @CRC_Research!