So you just did a round of user interviews or usability tests and need to make sense out of 250 individual quotes? Of course you know how to do semantic analysis with affinity diagrams or the KJ method. But how can you find out about the emotional side, how interviewees feel about the subject they talked about?
Here’s an easy, step-by-step method – check out whether it makes sense for you and your project. The basic idea is to create a tag cloud from emotional words that appear in the notes you want to analyze. The more frequently a word is used, the larger it appears in the tag cloud. The result is a display where the most-voiced emotions literally stand out.
As an example, let’s apply the method to the lyrics of Paul Simon’s song 50 Ways to Leave Your Lover. Did you ever wonder what the “friend” in the song was up to? Well, let’s see whether we can find out.
- You need the text to be analyzed in digital format – any word processing or spreadsheet format will do. That’s your text corpus.
- Build your list of words: you want to analyze emotions, so you need to sift out all words that carry emotional meaning – good, bad, elegant, clumsy, friend, threat, etc. Of course it’s an entirely subjective decision which words are emotional and which aren’t – trust your instinct, after all that’s what emotions are about, right?
The easiest way to build your word list is to delete all words from the text corpus that do not carry emotions in and by themselves. But watch out: you also want to correctly capture negations and emphasis, like “not good” or “very good”. Keep those, too. Better: change them to not_good or very_good, respectively; we’ll see shortly why.
- Standardize spelling. There may be 50 ways to spell the same word; you need to decide on a standard spelling. Just sort your word list alphabetically – related words then will show up next to each other, and you can change the differing spellings to a standard one. (Here’s my standardized word list for 50 Ways).
- Copy-paste your standardized word list into a tag cloud generator. There are several available for free on the web. I picked worditout because of a specific feature: if you enter something like very_good, worditout counts this phrase as one word, but omits the underline character in the tag cloud. So you have an easy way to keep phrases together without screwing up the display.
It took me some 20 minutes to go through this process for 50 Ways to Leave Your Lover, including googling the lyrics. Here’s the result (Figure 1):
Figure 1: Tag cloud of emotional words used in Paul Simon’s song 50 Ways to Leave Your Lover
Love shows up all right, but way more prominent are words about freedom, independence, change, and personal assertiveness. Think about it: isn’t this exactly the spirit of the 70’s? Maybe that’s what made the song so popular.
To be sure, this is not a method for big-scale data analysis. Manually going through the text corpus becomes tedious when you have more than 500 notes to go through, which however is a typical scale for user interviews or post-usability test interviews. Nor is it particularly accurate: you take samples out of a sample of utterances you captured from a sample of subjects in a particular sample of possible usage situations – you definitely need to take the results with a grain of salt.
This said, I find this a handy method to communicate in one slide the emotional essence of what has been said. I wouldn’t bet the farm on it, but quite possibly my next research results presentation.