Lab Week #3

I believe that textual data analysis is a very method-heavy projects. From my work with Voyant Tools, I found that its interface and its ability to create an interactive space filled with common words and phrases far outmatched Google Gemini’s generation ability. For example, even as simple as a word cloud was difficult for Gemini to correctly do without completely making incoherent words in the image. Compared to Voyant’s interactive cloud, which allows for a much cleaner and coherent model.

There is certainly some overlap between using the two tools. However, some of the mistakes make me at odds when using Gemini. The hiccups like “heav” and the one big “n” make it seem like it struggles with Old English writing. For context, I gave both tools a text file of Paradise Lost by John Milton, which is a familiar text to me. There were certainly fewer mistakes than I expected using generative AI. That being said, however, the discrepancies between the Voyant analysis and the Gemini one suggest that we should be wary of AI’s limited capabilities. Gemini proved that although okay for the textual analysis, it should not be relied upon as the primary tool. In fact, what you see is the second result of my prompting of Gemini. I decided to ask, “In the word cloud, make sure you are being accurate and applying stopwords?” after the initial prompt of word cloud, while excluding common words such as the, and, like, etc. The result is a good step in the right direction, but still not accurate enough to be relied upon for the heavy text data analysis. I will say regarding understanding the context of the text files, Gemini is much more easily able to give context on what the themes, narrative, and other surrounding information on what is in the text file. Because of that reason, I think it’s best if you have not read a particular file, to use Gemini to give you that context, then run it through Voyant to get the textual data.