Grading note:
	I'm caught up, in general
	Homework 5 was the funnest assignment I've ever graded

Text processing, parsing, and linguistics:
	I took a class on that once...
	Tokenization!

Consider a pretty cut-and-dried problem:
	Parsing a C-family programming language
	1.  Tokenization of input 
	2.  Forming a parse tree
	3.  If we were actually compiling, optimization
	4.  If we were actually compiling, generating machine code

Parsing English is a lot harder
	The C-family language was designed for easy parsing!
	English is not:
		The same set of letters can have more than one meaning
		The same meaning can be expressed by multiple words
		Speakers may invent new meanings for words or use local idioms
		Sentiment of the speaker, such as sarcasm, may change meaning
	
If you really want to learn to do this:
	https://explosion.ai/blog/parsing-english-in-python

When we read, we don't sound out words
	We don't even read the words
	All we do is pick from the sensible choices
	Same thing with listening

There's a lot of knowledge that goes into understanding English
	That's part of what makes models like ChatGPT better
	Imagine one could make a "sensibility" score
		Choose the most sensible interpretation
		Even if it's not the most technically correct

Prolog and first-order logic:
	Prolog demo

So, if we could extract facts, we could use first-order logic!
	An early idea
	Not really flexible enough
	We need to know how people really speak