I was originally curious to use Voyant on songs by Joni Mitchell because they are texts that I have a strong emotional response to. Since I am much more emotionally than clinically motivated with academic work, I was curious to use a tool that felt so cold on a text I feel so feverish about, and to see if it would produce any feeling or excitement in me stronger than, huh, that’s interesting.

For my sample I used the lyrics from Blue (1971), Court and Spark (1974), and Hejira (1976). They’re among her most critically and commercially acclaimed, as well as being ones that I happen to have an intimate familiarity with. I also liked, for no real reason other than shape, that there was another album released in between all of them.

The most commonly used words are not very interesting, although they are perhaps a good lesson in the geography of Mitchell’s music which is personal (I’m 55 times), descriptive (like 54 times), qualifying (just 43 times), and romantic (love 39 times). There are also words whose high counts are initially striking, but which mostly appear due to the repetitious nature of songwriting. I was initially struck by the frequent occurrence of green, before I thought to check and see how many of those instances came from the song Little Green (9 out of 11). If Mitchell were more prone to choruses instead of the single-line refrains she gravitates towards, I imagine this would be even more noticeable.

Most immediately striking to me was the high count for oh (37 times), an utterance that is potentially not a part of the author’s written lyrics, and is instead a translation on the part of the transcriber. These ohs illuminate the potential problems of text mining, specifically the lack of easily represented content, but they are also illustrative of the ways this tool – perhaps accidentally – has the potential to shine a spotlight on something that may have otherwise gone unnoticed. If I had been asked to pick a word to represent these albums I may have chosen sad (5 times), ambivalent (0 times), or aching (0 times), but oh embraces all of these terms without excluding joy (“oh love can be so sweet”), humor (“oh you’re a mean old Daddy”), or the refrain of wistfulness that runs through River (“oh I wish, 5 times). It’s not so much as a placeholder between the words, an exhale of yearning, but it was a sweet jolt of recognition to see it writ so large in Voyant’s word cloud.

Interestingly, what I found more meaning in than the frequently repeating words was the long list of words that were only used once or twice. These are mostly nouns and adjectives, and taken together it is their immense variation, their lack of frequency, that best illustrates Mitchell’s versatility and specificity as a songwriter. There is a pleasure in scrolling through the list and coming across the word pachyderm, for instance, and remembering that she used it as a bizarre euphemism in Blue Hotel Room. The context is still missing. Vain (1 use) is not particularly exciting to me, but used in combination with darling (4 uses) it transforms both words into something more intimate and specific. Perhaps the ability to identify word-types or parts of speech would help with this. If you could search for all nouns with their modifiers, for instance. But even knowing what was missing, I was surprised at the strong emotional response I had just to looking at a list of words in alphabetical order, and at the strong sense of person I got from beautiful stretches like shanty, shadows, shades, sex, settled or pushed, punishing, punched (all 1 use each).

I think the next step that I would take with this exploration would be to widen the scope to other artists, and see if the same patterns emerged, the same things felt interesting to me. How would the early work of John Lennon look, with its much more formal genre constraints? What about a writer like Taylor Swift, who blends the personal and observational qualities that Mitchell uses within a pop-music structure?