ROUGE-L for Summaries

ROUGE-L scores summaries by their longest common subsequence of tokens with a reference. It rewards getting words in the right order without requiring them to be adjacent.

With LCS length $L$, candidate length $m$ and reference length $n$:

Tokenise by lowercasing and splitting on whitespace. Implement lcs_length(a, b) in $O(|a| \cdot |b|)$ time with $O(\min(|a|, |b|))$ memory, and rouge_l(candidate, reference) returning {"precision", "recall", "f1"}. Empty inputs give all zeros.