Finding Patterns in Your Text with Voyant Tools

Voyant Tools is a text analysis platform that allows you to analyze a single text document or compare multiple documents. It is useful to analyze big [language-based] data over time.

Uploading:

Single Text Document: Paste the text into the text box. This prevents additional text from a webpage being included in your analysis. For example, if I want to look at the language in the Constitution, I should paste the plain text Constitution into the text box rather than the National Archives Constitution Transcript webpage to prevent “archives” appearing as a top word in my analysis, as seen below.

Multiple Webpages: Paste the links to each webpage in the textbox with each URL on a separate line.

Multiple Documents: Move all the documents you want to analyze into a shared folder for easy access. Click “Open” under the textbox and select all the documents you want to analyze. You have to select everything you want to include at once, as you cannot add more documents at a later time. This is why it is easiest to put your materials in one central location.

The Main Tools:

Cirrus: Essentially a word cloud. Check out Ike’s blog post on Wordle for more information on how to use word clouds in your research.

Reader: Reader allows you to view the full uploaded text and select individual words to see reveal frequency data across the text(s). For example, I uploaded transcripts for three oral histories I conducted on abortion last semester to analyze how much each narrator talked about health insurance in their interview. To do this, I clicked on the word “insurance” in the Reader section and Voyant Tools generated the below data.

The data shows that first narrator talked about insurance during one segment of the interview, the second discussed insurance throughout the interview, and the third narrator did not directly talk about insurance during her interview. This data can be seen in three different formats. In the top left section, you can see where in the interview the word insurance was used. In the top right section, you can see the frequency of use, and in the bottom section there are links to where in the text insurance is used to provide context for your analysis.

While this data provides little insight on its own, it is useful as evidence with the context of the interviews. Since I conducted the interviews, I know each narrator has a different understanding of her abortion—Narrator 1 was concerned with her own experience and aware of her privilege, Narrator 2 was concerned about women’s access to abortion who were less privileged than her, and Narrator 3 was still grappling with her relationships to her former partners and parents. Given this knowledge, I can use the frequency of the term “insurance” in each interview as a way to show that each woman conceptualized her experience with abortion differently, but without that knowledge, there is little direct meaning in the data.

Trends: The trends function allows you to compare the relative frequency (useful for comparing texts of varied lengths) of words across a singular text or between multiple texts in various graph forms. It can help show variation between the focus of differents texts or how language has changed over time.

Conclusions:

I found Voyant Tools to be unwieldy. It gave me a lot of data that I did not need or that did not have much relevance to any analysis, and it was hard to sort through the data to find meaningful relationships between different texts. Perhaps this is because I was comparing only 2-4 relatively short texts at a time, rather than large inputs of data that the platform seems intended for. Or it is because I did not know what I was looking for and I was hoping it would reveal interesting relationships between different texts. However, that does not mean it is not useful to any historical research.

Voyant Tools is exactly what it sounds like—a tool for historical research. It is useful to help bolster an argument or provide evidence for an argument, but it does not make an argument for you. A researcher has to know what they are looking for and have a deep understanding of their subject to use Voyant Tools in a truly productive manner. But, if that is done, it can help provide visual or numerical evidence for a historical argument.

Voyant Tools created a great “Getting Started” guide that I recommend exploring if you want to use Voyant Tools in your research. It has helpful details on the plethora of tools available on the platform, as well as a guide on how to save your corpus for future reference.

What do you think of Voyant Tools? Do you think you could use Voyant Tools in your research? Did you have any success using Voyant Tools yourself?

One Reply to “Finding Patterns in Your Text with Voyant Tools”

  1. Thank you, Emily! At first, I was a little bit confused by it because it was unclear where “Home” is and what would happen, if I refreshed the page. But overall, it seemed to me as a tool much more sophisticated than Wordle. I think Martin Jessop makes a good distinction between digital visualization and illustration: “The term ‘Illustration’ implies an image which serves only to support written language; thus the main carrier of information is the associated text not the image.” And I could see how I could use it to illustrate my larger point.

Leave a Reply

Your email address will not be published.