Hello everyone! For this week I covered readings 3 (part 1, part 2) and 5. In “Space, Nation and the Triumph of Region: A New View of the World from Houston” as well as its accompanying website, “Mining and Mapping the Production of Space”, Cameron Blevins explores history in relation to space and place. He demonstrates the importance of accompanying digital work with traditional readings to gain context in a more in-depth analysis. Blevins examines how the Houston Daily Post newspaper created an “imagined geography” from 1894 to 1901. Newspapers privileged certain places over others in their publications, which Blevins argues actively rather than passively “reinforced the structural power of capital” (136). Though Blevins’ methods for exploring this topic are what I think is most fascinating and important for class discussion. Blevins utilizes “distant reading” as a method. It would take too long to read all the newspaper articles, he estimated three years to do so. So instead allowing a computer to pull out keywords, like the number of times a place is named in newspaper articles, the time it takes to make a general analysis of the “imagined geography” created by the Houston Daily Post can be achieved under a shorter amount of time. This provides a better visualization, which is key for spatial history analysis. Though this paper reminds us in its conclusion that such a preliminary analysis rarely tells the whole story and it needs to be accompanied with “close reading” of traditional texts in order to get a better understanding of the context. For example what was discovered was that even though the preliminary search suggested that there was a national scale oriented towards New York and the Midwest as well as the dominant regional scale of Texas and its immediate orbit, this did not reflect most populous cities as assumed at first glance. Rather with further textual analysis and close reading, Blevins discovered that this newspaper’s focus centered on commercial content like cattle, wheat flow, cotton and finance. Moreover, the Houston Daily Post revealed “an unexpected prominence of regional space” (137). This, Blevins discovered was due in part to shifts in American journalism around that time. Newspapers were becoming highly commercialized entities targeting a mass-market of readers. As a result, The Houston Daily Post appealed to its most reachable audience within the local region. Another point made by Blevins was that the railroad networks seem to mirror the most common places named in the newspaper. This is also misleading because after further examination through close reading, how often a place was named was related to the flow of goods and capital as mentioned above rather than railroad hubs or large populations. Blevins’s accompanying website offers interactive maps of his findings that allows readers to further explore his methods and the importance of critical analysis. Distant reading is great to see an overarching trend but in order to make a claim about that trend further analysis is required. As he summarizes his work he notes how we need historians to be aware of the need for close reading and text analysis more than ever with our constant use of the internet. The easily available and potentially misinterpreted “distant reading” that is abundantly on the web calls for critical approaches to history and the addition of close reading to get the whole story.
This theme of textual analysis and distant reading is carried over into Cameron Blevins’s other article, “Topic Modeling Martha Ballard’s Diary”. Blevins presents the difficulty of reading and analyzing an extensive diary, similar to the extensive newspaper articles. Through the use of MALLET, a computer software that attempts to find words that frequently appeared together in texts and group them together within clusters. This is called text mining or classifying words from text based on assumed sentiment, intent and topic. Moreover, topic modeling is a text mining tool used to pull out semantic structures from within a text. Blevins found that this program is actually very efficient at pulling out a handful of themes from this diary. Midwifery, church, death, gardening, shopping, and illness were the big themes that this program pulled out. Now how accurate is this? Well, Blevins tested it by examining some further close reading to determine if this overarching analysis was accurate. For example, the appearance of “gardening” and words that the program determined were used similarly or related to that word did follow the pattern of Maine’s seasonal cycles, which is where Ballard lived. Furthermore, the mention of housework increased over the course of Ballard’s life which correlated with changes in her history that are known like her children moving out as well as increased financial troubles which would definitely cause her to spend an increasing amount of time doing housework. Blevins further demonstrates the benefits of distant reading that MALLET can provide but it needs to be accompanied with close reading to get the better visualization of whole story. As usual context is everything. There are so many possibilities with topic modeling. Using it across languages or on social media and blog posts were particularly interesting questions and hopeful suggestions made by commenters. Rather than hesitation and suspicion though, Blevins tries to show that the possibilities are truly endless and these computer programs have opened up an entirely new world for history. But it’s important to remember that just because we have these programs doesn’t mean we can discard close readings but instead that they are key in supporting the evidence that these programs provide us.
As practicing historians do you consider Blevins approach beneficial or ideal in the face of constantly advancing technology? We have no way of knowing how perfected these computer programs, or others, may become, but do you recognize the value in the method of combining close reading alongside distant reading?
Thanks everyone keep up the good work!
2 Replies to “Using Technology to Visualize the Past”
Hi Claire! I thought that the various digital tools that Blevins used to show his research were really interesting, especially with Martha Ballard’s diary. I was also impressed with MALLET’s ability to pull data from the diary entries by tracking patterns and how the words were used in the text. It was especially interesting to see how the program could find patterns around the theme “emotion.”
For accuracy, I feel like Blevins makes a good point about the impact of big data for finding the themes. Since Martha Ballard’s diary included daily entries for 27 years, there does seem to be a smaller margin of error since the program combs through so much information. At this point, maybe it depends on the type of conclusions that the researcher is trying to draw from the information. It kind of feels like the way that traditional primary sources require additional context for us to draw conclusions or analyses.
Yes I completely agree. Plus context is everything, especially when it comes to history so having reliable primary sources to support/confirm computer programs accuracy can only be beneficial. I also thought that MALLET’s ability to detect emotion through words and similar usage of other words blew my mind. So cool! and so much potential for other fields.