2017-02-15 - Questions 8 (Dylan)
Textual Analysis Readings
The readings for this week all addressed a common topic: The applications of textual analysis, specifically stylistics. In the Burrows reading, we get a few examples of how quantitative text analysis can help us answer questions about texts, but also raise new ones. This analysis was mostly stylistic, focusing on counts of certain words across several texts. This reminded me of Franco Moretti’s ideas about a quantitative theory of genre. However, the Burrows reading approach seemed to encourage the use of text analysis not to make sweeping claims, but rather to view texts from a different perspective and use quantitative textual analysis in tandem with interpretative work. This mixed approach was also emphasized in the Sinclair and Rockwell reading, in the idea of the “agile interpretive cycle:”
“We read texts we enjoy, we then explore and study them with analytic tools and visualization interfaces, which then brings us back to rereading the texts differently.”
This somewhat addresses the problem some humanists have with Franco Moretti’s ideas, namely that there’s a lack of real meaning and purpose as human interpretation takes a backseat to computer processes. We require both the speed and scale of Moretti’s distant reading, as well as the specificity and insight of close reading.
That’s all easily said, but implementing it is tough. When you want to do a project quickly, how do you ensure the data is up to the standards of close reading?
In Linguistics 1 last quarter, we did a study of a collection of data from conversations in Palo Alto cafés. However, since our data wasn’t correctly formatted in a lot of cases, and we didn’t have time to go through and correct all of it, there were some inaccuracies that slipped through the cracks.
The article on Google’s NGram viewer mentioned in the Sinclair and Rockwell reading also mentions some potential pitfalls. What resources/strategies exist to tackle these problems?