Freeform

The tokenizer in this little text lib is producing bad output: punctuation rides along with the words, so "Hello, world!" comes back as ["hello,", "world!"] instead of ["hello", "world"]. The bug is in src/parser.js — the split there only breaks on whitespace, which is why punctuation never gets stripped. Fix it so tokens come out as clean lowercase words; apostrophes inside a word (like "don't") should stay put. The suite is failing right now — run `node --test` and it should pass once tokenization is right.