In Which I Am Personally Attacked by the Topic Model#
Once again, here’s what I expected to find in this corpus before I started investigating:
Basketball
Family
Music
Mental Health
Gender Stuff
School
And here’s what one round of asking Topic Modeling Tool for 10 topics gave me. (Here, I’ve entirely removed the names of people I personally know that show up here. I also crossed out instances of partial contractions — what? — as well as partial numbers, e.g. “000,” since this probably appeared following a comma as part of a whole number like “1,000” rather than as “000” alone.)
Topics#
Topics |
Brief Interpretation |
|---|---|
1. drinking thing 30 alcohol wondered days tiktok normal pandemic counting panic drinker piece burnham scary worse shopping decided uncertainty |
COVID-19 pandemic, during which I intentionally decided to quit drinking, watched a lot of TikTok, turned 30, and was generally uncertain |
2. students school room day challenge soccer multipurpose parents line replied sixth looked cross tears entire change hand university |
K-12 education and the transition to college |
3. family mom haiti haitian country history paper bread duvalier born california ms texas french uncle dad spokane haitians united |
Reflections on Haitian family history |
4. anxiety bed disability dr student room thoughts night ward roommate energy safe death class academic leave spoon darkness |
Mental illness history |
5. mom sonoma parents sister fire town text read smoke alert department sheriff safe air guilty evacuation evacuations neighborhood frosting |
Following 2017 Northern California wildfires from afar |
6. father body mother weight parents world feel reason anger person social quarantine numbers physical emotions fact public focus play club |
Growing up with my parents |
7. school band music class high ms teacher flute violin instrument year concert years mrs play students playing |
My history playing music |
8. pronouns easier woman people gender shirt binary open red posted writers cis white door suicide pronoun nonbinary |
Considerations of gender, pronouns, being nonbinary |
9. gonzaga basketball women team game games band tournament campus men zags arena season ncaa bed schedule win pep spokane roommate |
My experience supporting the Gonzaga women’s basketball team |
10. time people day school year life |
History of my platonic relationships |
(There’s just something about a tool you trained with your own writing telling you what it “thinks” of you.)
My most significant past experience with topic modeling was a group project in a DH class. We looked at song lyrics and lyricists of mid-century musicals, focusing both on the content of these songs and whether it was possible to identify a lyricist through the lyrics alone. I learned a lot about the subjectivity of topic modeling — how a tool can offer its take, but it’s up to you to extract meaning.
But here, of course, this process could not get more subjective. My friends’ names appear as topics. I can trace certain storylines linearly, but it also combines aspects of my life that make me go, “Wait, these things go together? Huh, I guess these things go together.” Everything I expected to be represented, is represented.
But the subjectivity feels somehow more urgent here, especially when the topics seem vague. Some of these outputs feel obvious to me. For instance, Topic 8 mentions pronouns, gender, binary, cis, and woman — a lot of my writing around the time I came out as nonbinary was about navigating gender stuff, as my own summary at the top of this section suggested. Meanwhile, Topic 3 feels similarly cut-and-dry: all of the topics about Haiti are clumped together, and my Haitian mother is mentioned. I do think there’s something to say about how 3 of the 10 topics explore different aspects of family, but this could be because I can find many avenues to mention them, so why not the tool?
Topic 9, though, was where I first felt the need to intervene. This topic pretty exclusively traces my history supporting the women’s basketball team at Gonzaga University, where I started undergrad. Topic Modeling Tool drew from 9 of the 37 pieces in the corpus to create this topic, including 947 words from one in particular (the second-most words from a single piece for this topic was 81). That much makes sense! It was a long piece that mentioned Gonzaga basketball the most. However, the piece is about my relationship with Gonzaga basketball in the midst of worsening mental health crises. That Topic Modeling Tool opted to separate these felt notable — or, was it just a reflection of my writing that I didn’t do enough in the piece to connect the two?
In the spirit of crowdsourcing in DH, I asked two close friends with DH experience (i.e., who are familiar with topic modeling) to share their immediate takeaways from each topic. I was interested both in the strength of the topics — would someone who knows me, but who isn’t me, be able to extract similar meanings from them? — and whether they might choose to elevate words I might not in their pursuits of forming a topic.
Topics (Revisited)#
Topics |
Brief Interpretation |
Friend A’s Interpretation |
Friend B’s Interpretation |
|---|---|---|---|
1. drinking thing 30 alcohol wondered days tiktok normal pandemic counting panic drinker piece burnham scary worse shopping decided uncertainty |
COVID-19 pandemic, during which I intentionally decided to quit drinking, watched a lot of TikTok, turned 30, and was generally uncertain |
getting sober during COVID lockdown |
anxiety and its causes and coping |
2. students school room day challenge soccer multipurpose parents line replied sixth looked cross tears entire change hand university |
K-12 education and the transition to college |
being at school |
uhhhhhhhh this one is hard. change? growing up? idk |
3. family mom haiti haitian country history paper bread duvalier born california ms texas french uncle dad spokane haitians united |
Reflections on Haitian family history |
family history |
🇭🇹 |
4. anxiety bed disability dr student room thoughts night ward roommate energy safe death class academic leave spoon darkness |
Mental illness history |
body and mind struggles |
disability |
5. mom sonoma parents sister fire town text read smoke alert department sheriff safe air guilty evacuation evacuations neighborhood frosting |
Following 2017 Northern California wildfires from afar |
family and a fire |
disaster of some sort |
6. father body mother weight parents world feel reason anger person social quarantine numbers physical emotions fact public focus play club |
Growing up with my parents |
family and your body/health |
existing in the world in a body is hard |
7. school band music class high ms teacher flute violin instrument year concert years mrs play students playing |
My history playing music |
school and band |
muse ick (sic) |
8. pronouns easier woman people gender shirt binary open red posted writers cis white door suicide pronoun nonbinary |
Considerations of gender, pronouns, being nonbinary |
gender and mind |
Gender Trouble™ |
9. gonzaga basketball women team game games band tournament campus men zags arena season ncaa bed schedule win pep spokane roommate |
My experience supporting the Gonzaga women’s basketball team |
sports |
sport |
10. time people day school year life |
History of my platonic relationships |
nostalgia or some other kind of relationship to the past |
autism things tbQh |
I know this project is about my writing, but I have to mention how fascinating this exercise was in terms of learning how what I’ve told my friends — who haven’t read all of this writing — represents itself in their interpretations of these topics. I’m introducing unnecessary variables, I know! Sue me!
While I knew how some topics misrepresented parts of my life, as in Topic 9 highlighting Gonzaga basketball without the context of mental health, I was very interested to hear my friends’ interpretations about two of the more nebulous topics where I felt I reached the most to make meaning: Topic 6 and Topic 10.
6. father body mother weight parents world feel reason anger person social quarantine numbers physical emotions fact public focus play club
I tagged Topic 6 as “growing up with my parents,” which I think is the ultimate form of disengagement and distilling a lot of complex feelings into a hand wave, a curt acknowledgment of things that may have happened that I don’t wish to revisit in this medium. But I chose to ignore terms like “quarantine” and “club” that didn’t fit this interpretation. My friends, on the other hand, went with “family and your body/health” and “existing in the world in a body is hard.” And neither of these are incorrect! Neither of those don’t take into account that my familial experiences overlapped with the aspects that they chose to highlight. But while my takeaway was “haha classic family,” theirs delved more into the explicit elements of the topic that I was able to summarize for myself with “well clearly, those things are family-related.”
As mentioned previously, Topic Modeling Tool allows you to see which documents factored most into the topics. Three of the top four (each of which had between 150 and 400 words that factored into the topic) were essays I wrote in undergrad for a freshman comp class. All three were specifically about my family. The fourth was a reflection on the “quarantine 15,” which doesn’t not figure into comments I received about my own body growing up.
10. time people day school year life didn back felt make knew friends don made things home wasn feel friend work
Topic 10 felt like a trash can, honestly.[1] All three contraction fragments ended up here, and what remained were a bunch of generic-seeming words that felt like they applied to a person in general, not me specifically. Time? It passes. School? People typically go to it. Friend/friends? A thing people tend to have. “History of my platonic relationships” arose from my desire to connect the dots by force — I tend to have made my friends at school or work, for instance. And they are…people? In my life? Who I feel things about?
I just really appreciated the earnest attempt on Friend A’s part to make sense of this one — “nostalgia or some other kind of relationship to the past” (which, again, is not untrue) — while Friend B went full vibes with “autism things tbQh” (we are both autistic). Both are valid.
In conclusion, the topic model doesn’t know better than me, but I feel like I have a better understanding of what a distant read prioritizes. Now, let’s ruin it!