For this project, I worked with The People of Palestine by Elihu Grant, a public-domain text from Project Gutenberg. I analyzed it using two different computational tools: Voyant Tools and Google’s Gemini AI. Using both made it clear that different tools “read” the same text in very different ways, and that difference really matters, especially with how much we rely on AI right now.
I started with Voyant Tools to get a big-picture sense of the text. The Cirrus word cloud immediately highlighted words like “village,” “country,” “Allah,” and “footnote.” Even without digging deeper, this showed that the book focuses heavily on rural life, religion, and a more academic, observational style of writing. Voyant was useful because it revealed patterns across the entire book very quickly, without needing to read it closely page by page. At the same time, it has clear limits. Seeing that “Allah” appears frequently tells us religion is important, but it does not explain how religion is experienced or discussed. Voyant is great for spotting patterns, but it cannot explain meaning in and of itself.
For the AI portion of this project, I followed Anastasia Salter’s distant reading workflow and uploaded the full Project Gutenberg text of The People of Palestine into Google Gemini, as instructed in the tutorial. I asked the AI to generate a word cloud of the top 30 most frequent words in the text. This showed how AI attempts to perform distant reading at scale, but it also made its weaknesses obvious. Compared to Voyant’s word cloud, Gemini’s output felt less transparent, especially in terms of how words were filtered or prioritized and what choices were made behind the scenes.
Working with both tools made it clear that AI can be useful, but only up to a point. The AI-generated results looked clean and convincing, but they simplified a text that is culturally and historically complex. Comparing Gemini with Voyant helped me see why computational tools work best when used alongside other methods and human judgment, rather than when relied on alone.
I really like the point you make about having to use judgment with these computational tools like Gemini or Voyant Tools. I had a similar position that I wrote about in my lab assignment, and I totally agree. Additionally, I think it’s really interesting how Gemini constructed your visualization of the top 30 most frequent significant words. The style that the AI implemented with my word cloud was vastly different. Lastly, I like your thought about the Gemini output feeling a lot less transparent. Although AI can be really helpful, it can still make mistakes and incorrectly give us information. Great points!