16 · Text I — regex

Tuesday, Oct 20, 2026

Materials for this session are not published yet. They appear here before class.

Objectives

By the end of this session you can:

• Read and write practical regular expressions.
• Extract structured fields from messy text.
• Pair every pattern with positive and negative tests.
• Judge when a regex beats a model, and when it doesn’t.

What we cover

• What text has measured: media slant, FOMC language, skills in job ads — the text-as-data classics.
• stringr; regex from literals to character classes, quantifiers, anchors, and groups.
• Counting words: tokenize, table(), stop words, term frequencies — the dictionary tier that answered those classic papers, and still the right first move on any corpus. Readability metrics in one slide; fixing a broken character encoding in one line.
• Regex and counting as the cheap, deterministic tools you try before reaching for a model — session 17 is for what they can’t do.

Verification habit. Every pattern ships with a should-match test and a should-NOT-match test.