The other half of studying language.
Studying how machines read language is one half of the interest. Writing it myself is the other. Reading closely is how I noticed the problem CNM-BERT exists to solve.
Recognised work
Three Gold Keys in the Scholastic Art & Writing Awards, and a quarterfinal at the Harvard World Schools Debate Open.
Argument is a technical skill.
Debate taught me to find the weakest joint in my own case before someone else does. It is why the research pages here state each result's limits next to the result. A reviewer finds the gaps anyway; better they come from me.
The reading habit is less disciplined and probably more useful: world literature, philosophy, and theology, mostly outside anything I am formally studying. CNM-BERT exists because I found it strange that a model could read a million Chinese sentences and still know nothing about how a character is put together. I got there from the reading side of the desk, which is probably why it looked strange to me rather than routine.