I chose to use The Great Gatsby for this assignment. First, I entered it into Voyant Tools. I really appreciated the interface of Voyant Tools. Each aspect feels interactive, and I liked being able to look at multiple visualizations for different or similar data at the same time. I also think they have a good number of options for different types of visualizations, which makes it really useful for various purposes.
The one I included here was the “Dreamscape” which I think is a really cool feature that pulls geographic words/town names out of the text, and not only maps them but includes the frequency at which they’re used. It’s not the most accurate, as we saw with the Shakespeare example that assumed any of the town names were referring to the American counterparts of the English towns that Shakespeare actually referred to. This one seems a bit more accurate, possibly because the locations are mostly all in the US. It’s a fun and unique way to visualize the information from a novel, though geospatial movement might not be the most important aspect of The Great Gatsby.
After Voyant Tools, I used Gemini for data visualization. This I actually really struggled with. At first, despite following Anastasia Stalter’s instructions, it just estimated word frequency (see below). Even after prompting it to use actual numbers from the txt file I attached, it said it could not, and instead provided me with Python code to count word frequency in a txt file. Though this seemed somewhat helpful, it was frustrating that it did not actually give me any text analysis when I asked for it. In terms of generating a word cloud, it gave me fabricated data, and when I tried to run the Python script that it had provided, I got a ton of errors, and it did not work. This surprised me, as in the past I’ve felt writing/working with code is something AI is most effective at.
Additionally, when I strayed from Stalter’s direction and tried to get Gemini to generate images of the text analysis, it produced images that were essentially nonsense. I tried to get it to be as factual as possible, but it just did not want to.
Overall, I found a lot of difficulties using AI, whereas using Voyant Tools was a lot more straightforward and easy. I suppose Gemini might just not be programmed in a way that lends itself to text analysis, and maybe there is an LLM that’s more effective. I also wouldn’t rule out the possibility that I was just making mistakes, but I did try my best. In terms of the “Age of AI”, I would argue people should use it as a tool rather than a shortcut, and that it has a lot of possible benefits. However, those benefits are not showcased by my experience. There needs to be a lot of double- and triple-checking when it comes to AI, or else you end up with word clouds based on entirely fabricated numbers.