Accent and Sentiment
January 6, 2016 7:38 AM   Subscribe

Two recent essays explore the way our expectations and our language interact in fiction: First, at The Toast, Brittany K. Allen deals with what "urban" means in "urban romance," and how hewing to the genre affected her writing, in "I Wrote the Accent." Next, Andrew Piper and Richard Jean So use sentiment analysis to determine whether pop fiction is more sentimental than its literary cousin, in "Quantifying the Weepy Bestseller."
posted by mittens (5 comments total) 14 users marked this as a favorite
Is it obvious that sentiment analysis is going to be valid across datasets with systematic differences in style? That seems like an open question to me. Haven't read the piece yet.
posted by grobstein at 12:11 PM on January 6, 2016

Is it obvious that sentiment analysis is going to be valid across datasets with systematic differences in style?

To answer my own question: no, you probably can't compare sentiment robustly this way. What you can compare is the "resultant" of sentiment and style differences. So S_1 = "Hitler was a horrible guy" and S_2 = "Hitler eliminated a whole generation" might express the same underlying sentiment, but with different styles. For any given sentiment model, S_1 will probably be measured as more strongly negative than S_2. Since style varies systematically between the comparison groups, we can't tell which differences are due to underlying sentiment and which are due to different style.

But this is still an interesting regularity to study.
posted by grobstein at 12:25 PM on January 6, 2016 [1 favorite]

I liked this and am glad you posted it. Thanks.
posted by not that girl at 3:19 PM on January 6, 2016

I'm delighted and fascinated by these as individual pieces, and by the juxtaposition.
You have made me a happy book nerd, and also a happy librarian. Thank you, Mittens!
posted by SaharaRose at 3:24 PM on January 6, 2016

The first essay is excellent (and enraging: seriously, telling African-American writers they're "not black enough"?); the second has interesting data, but the authors should have confined themselves to interpreting it in terms they understand and skipped the condescending and pointless advice to authors ("just write a good novel": sure thing, boss!). Thanks for the post!
posted by languagehat at 8:40 AM on January 7, 2016

« Older Les Guerriers de l’Ombre   |   "When Your Fat Pic Goes Viral as a Feminist... Newer »

This thread has been archived and is closed to new comments