Machine Learning and Human Expertise in the Slavic Digital Humanities

Andrew Janco

Haverford College 

Humanities at Super-Human Scale

  • Distant reading 
  • Distant Horizons

 

Able to study more than a set "canon" of works

 

Ability to analize more materials can could be studies by a single scholar in a lifetime.

 

 

🦜 Stochastic Parrots and Bias at Scale

Katherine Bode

 

Emily Bender

Timnit Gebru

Micro-DH

John Unsworth "Scholarly Primitives" (2000)

 

“basic functions common to scholarly activity across disciplines, over time, and independent of theoretical orientation.”

annotation and association

Text Markup and TEI

Annotation /  Named Entity Recognition

🗑️garbage

not 🗑️garbage

Prozhito 

Association / Named entity linking 

  • Start with "scholarly primitives," not scale of analysis or size of datasets
  • Rather than appropriating innovations from computer science and artificial intelligence, this approach builds from humanities-based methods and practices.
  • This view shifts our focus from the potential intelligence of machines towards their ability to complete useful tasks for human ends. Specifically, how they can augment scholars’ work by performing repetitive tasks at scale with superhuman speed and precision.
  • By augmenting these activities, we are able to benefit from the scale and precision afforded by computational methods, as well as the valuable interplay between scholars and machines as humanities research practices are made explicit and reproducible.  

Thank you!