The New Zealand Asian Studies Society 23rd Biennial International Conference Presentation (November 2019)

I presented the following slides at the NZASIA held in Wellington this past year. Because I had spoken about the results and significance of my research on previous occasions, I decided to do something different and discuss my methods and methodology. For the purposes of this talk, I focused on my second case study because it contained the largest number of primary sources. As such, this case presented many methodological challenges, and I would like to share how I attempted to overcome them.

Developing new tools and methods for approaching large sets of textual data can facilitate the study of historical writing with a transcultural component. When I say "transcultural data," I refer to the known and unknown ways that Chinese and Korean linguistic, cultural, scholarly and even musical practices brought about the songs in these 3 sources. [SLIDE 4]

The traditional method of analyzing would start with reading all the lyrics of every song of the NAUA. But imagine if there was a database of songs in both languages—how might we use methods from Big Data, like natural language processing (NLP), to analyze the songs in Chinese and Korean side by side? (I am new to this idea myself, but over at The Fish and the Painting, Professor Andrew Piper has given me some ideas on how to begin creating such tools).

I decided to build a database that I hoped would help me read the lyrics quickly and efficiently. I wanted a tool that would provide information about the songs regardless of language. In other words, before I did any close readings, I wanted to get a better sense of the patterns that linked the songs and whether those patterns held close to language or not. A lot of the information I was interested in was themes, imagery, phrases, and other associations between songs. I imagined that if there were a database, many associations would naturally appear through the use of various algorithms or search terms.

Sadly, I'm not a programmer, so I used NVivo to digitize the songs. My manual process of entering them into the software was, inevitably, a type of reading in itself (a not-so-distant reading, if you like). Nevertheless, it led to some important observations about the songs as a corpus. I was then able to use the in-built tools in NVivo to quantify and qualify some of those findings.  

[SLIDE 7 & SLIDE 8]

Inputting the songs and coding them resulted in 31 preliminary categories based on imagery, metaphor, titles and subject matter, specific words used, and themes. The files tell us in how many songs I coded a particular idea. The references demonstrate how many individual times an idea was coded, which means that I coded the same idea multiple times in the same song. Why was this coding process useful? 

[SLIDE 9]

It gave me some empirical data about my hypothesis, which was that Manchuria as a place was a prevalent idea for Chinese and Korean NAUA soldiers. Coding also led me to some ideas that I hadn't anticipated. 

[SLIDE 10]

For example, I had not expected the prevalence of flag imagery or how such a concept would be layered with other ideas. The color red was often paired with flag, which indicated that these terms were being used as symbols of communist ideology. On this slide, you’ll see two different ways of visualizing the layering of flag imagery and the color red in the songs. 1) You can see the songs that were coded into both categories. 2) This is a word cloud of the most frequent terms in the flag category. Interestingly enough, the four biggest characters we see are “red flag” in Chinese and Korean right in the center overlapping each other. To me, this indicates that the NAUA song corpus should be conceptualized transculturally. Certain ideas—images, if you will—are statistically given more weight than language. To end, I'll show a couple more visualizations that NVivo offers as a way to conceptualize this data.

[SLIDE 11]

One of the most prominent themes in the NAUA songs is unity. I quite like how this image captures the idea that unity lies at the center of not only a number of Chinese and Korean songs but also the songbooks from which they originated. Such an image suggests that there is a deeper connection between the NAUA and cultural representation as it occurred in northeast China after Manchukuo was dismantled. 

[SLIDE 12]

Finally, here is another word map created from the word frequency in the ideology category. Admittedly, I coded a lot of random words and phrases as ideology, so this word map is not the most accurate depiction of ideological currents per se. However, I'm showing this to you because, like with the previous unity image, the NAUA took great care not to let cultural divides or nationalistic political objectives get in the way of working together to reclaim Manchuria from the Japanese. As such, many of the larger characters in this image speak to the ways NAUA soldiers sang about their goals. For example, we have "gong" (work) close to "chan" (property) surrounding the Korean word "ro-dong ja" (laborers); then there's Lenin, class, war, country, "quan" (whole) next to "oo-ri" (we), and "oo-ri hyeong myeong" (our revolution), "dou" (everyone) above "kang" (resist). Despite this rather compact-looking word map, I'm sure can all agree that there are parts of it that don't make much sense. Yet this, too, is part of transcultural methods. The authors of Transcultural History: Theories, Methods, Sources state: "Focusing on entanglements, a transcultural approach always addresses the argument of culture in a deconstructionist way...a transcultural approach avoids the harmonious picture, mistrusts the idea of a harmonious coincidence, and focuses on the hair-line cracks that separate the pieces of the puzzle." I would like to leave you with that food for thought about, perhaps, your own future transcultural projects.