For this project, I used a public domain text from Project Gutenberg and two computational approaches to text analysis. I used Voyant Tools to look at word frequency and patterns, and Gemini to summarize the text and highlight the AI model’s process.
The first method I used was Voyant Tools, specifically the word cloud visualization. The largest words in the cloud, such as “Joel,” “west,” and “ball” were either very generic words, central to the story or names of characters in the story. It tells you right away that the main characters or at least most mentioned characters were Joel and Outfield West. You can also tell that it’s a sports book because ball is mentioned a lot. Voyant makes it easy to notice these patterns, which is helpful when working with longer texts or gaining some initial knowledge about a text. At the same time, the visualization shows one of Voyant’s biggest limitations. Words like “well,” “know,” and “right” appear frequently but do not carry much interpretive weight on their own. Voyant shows what appears often in the text, but it does not explain how or why those words matter. There is not a lot of context provided without further analysis.
To compare this approach, I used Gemini to generate a summary of the text by just simply asking “summarize this text” with the url included. It was able to summarize major themes, recurring ideas, and describe relationships between characters. This kind of analysis offers a quick look into the text and provides context for the reader before a deeper dive. I asked Gemini what its process for summarizing was which is highlighted in the image below. You can see how it first mistook the url for the book “Pride and Prejudice”. Then it pulled the context and broke the book into important lenses. This is useful, however it also raises concerns. AI often sounds confident even when its conclusions are vague or not clearly supported by specific passages. Without checking yourself, it would be easy to misinterpret the text if the AI model was wrong or taking something out of context.
In the Age of AI, one of the biggest concerns is how easily computational tools can replace close engagement with a text. While tools like Voyant and AI models are useful for spotting patterns and generating ideas, they should not be treated as final authorities. They work best as starting points that support, rather than replace, human reading and critical thinking.
I liked how you compared Voyant’s pattern spotting with Gemini’s summaries, especially the example where Gemini confused the text at first. That moment clearly shows why AI can sound confident but still be wrong. Good reminder that these tools work best as starting points, not final answers.