Using the Voyant text mining tools for the first time, I found the word frequency data and the corresponding visualizations (such as the Cirrus word cloud and the Trends graph) to be incredible valuable for gathering insights about the WPA Slave Narrative interviews. When done correctly, to include flagging common “stopwords” to filter out irrelevant data, an analysis of word frequency can prove fruitful for understanding the topics of emphasis for the authors of the corpus. And yet more than telling us the about the corpus as a whole, the distinctive words tool also makes an intra-text analysis possible – one in which the user can compare different sections of the corpus against each other. In the case of the WPA Slave Narrative documents, this tool revealed differences in dialects between interviewees of different states.

But as Sinclair and Rockwell remind us, these tools do not create inherent meaning from their results; determining the significance of the results still relies upon the researcher. For example, the Link tool revealed a relationship between the word “big” and the word “house.” Taking the results at face value would suggest that interviewees were simply referring to the physical size of the structures on the plantations they worked. However, as the student using the tool, it was up to me to connect this with the historical concept of the “big house,” which was embedded with cultural connotations connected to slave hierarchies and treatment on the plantation. Furthermore, as the Signs@40 activity demonstrates, our analysis is only as good as our corpus. In the process of selecting a corpus, whether from an archive or a journal in the case of Signs@40, we depend on the conscious or unconscious decisions of archivists and publishers make about what types of sources belong. Guldi refers to this as the “silence” of archives (523), and the Signs@40 project’s focus on the history of only one journal to serve as a stand-in for feminist discourse over several decades was one of its shortfalls.