Keywords and characters: Romeo and Juliet

This video-talk, though voiced by Jonathan Culpeper, incorporates the words of Sean Murphy.

This video-talk examines how a keyword analysis can reveal how language characterises Romeo and Juliet. It captures the symptoms of their linguistic styles or idiolects.

Focusing on all the talk of an individual character (something that is easy to do in CQPweb), the computer can extract the keywords for the character and rank order them in terms of how significant the keyword is. A further step one can do manually is to group the keywords that seem to be semantically related. For Juliet, for example, one obvious group is “swear”, “word” and “speak”, all of which seem to relate to language/communication. The point of this is to expose “themes” in a characters construction.

Keywords normally do one of two things. One is that they confirm what you always suspected. For example, the fact that Romeo has a group of positive keywords relating to romantic love (“love”, “beauty”, “fair”, and so on) is hardly surprising. But now we have empirical evidence for what we suspected, and that empirical evidence is precise. The other is that they reveal what you did not suspect. For example, the fact that one group for Juliet consists of “yet”, “if”, and “or”, and so on is not something one might have guessed. However, when one scrutinises the use of these keywords, it makes total sense: these keywords capture Juliet’s constant articulation of her anxieties.

Two warnings about the keyword results for characters are in order. One is that simply looking at a list of keywords as revealed by a computer is not enough. You need to go back to the text and look at how those keywords are actually being used in context (and this is why we show you examples in this video-talk). The other is that keywords vary in terms of dispersion. For example, one of Romeo’s keywords is “banish”, but that is confined to the scene after his banishment is announced. So, this keyword reflects local circumstances. In contrast, keywords which are fairly well dispersed (e.g. Juliet and the “yet”, “if”, etc. group) tend to reflect their character.

“How to” video-talks will be coming up shortly from Andrew Hardie. For now, if you have any thoughts – perhaps things that struck you as interesting in the results – or any queries, put them in the comments now.