The Observation Theory EncyclopediaFrom TSKAboutBy kindBy chapterBy Lean fileLedgerProvenance

ROUGE concept

DefinitionA family of scores for generated text against a reference. Chapter 0 section 0.11.
ExampleA generated text that reproduces 6 of a reference’s 8 unigrams has ROUGE-1 recall 0.75.
BookData Mining as Observation, draft 0.2, commit f3914f0; entry id rouge, kind concept.
Statusno ledger row names this entry. Corrections: none recorded.
Defining equationnone
Assumptions and scope
  • A family of scores for generated text against a reference. ROUGE-1 recall counts the reference’s words the candidate contains, with multiplicity, over the reference’s length. It lies in the unit interval, is one for the reference itself, and is one for any rearrangement of the reference, since it reads bags and not order.
  • A ROUGE score therefore certifies less than it seems to. The 13.7 ROUGE-L difference at 512 tokens re-validated at negative 0.31 on forty documents, and the claims ledger carries both numbers.
Prior artnone recorded
Evidenceturboquant-pro/CLAIMS.md:63-81, turboquant-pro/benchmarks/kvquant_matrix/REVAL-2026-08-08.md, lean/DataMiningAsObservation/Rouge.lean
Reviewednot yet reviewed; generated 2026-09-10 from records at the commits on the provenance page.
referencegeneratedreference ∩ generated
The overlap of the generated text's n-grams with the reference's.

Equation

none

Conditions

Conditions are curated in entries.toml rather than read from a record.

Ledger

none

First stated

Lin, ROUGE, 2004, as chapter 0 section 0.11 states it, with the program’s 13.7 ROUGE-L number and its re-validation in turboquant-pro/CLAIMS.md:63-81.

Measurements

Where the book states it Numbers, as the book’s sources table records them Source
chapter 8 section 8.9 13.7 ROUGE-L, 0.25/4.19/9.60/13.7 at 64/128/256/512, re-validation negative 0.31 n=40, 26.64 under symmetric nf4, _quant_nf4a_group unchanged since 289bdfc before 4f7baab turboquant-pro/CLAIMS.md:63-81; turboquant-pro/benchmarks/kvquant_matrix/REVAL-2026-08-08.md

Failures and corrections

none

Invariance envelope

none declared

Machine checked

lean/DataMiningAsObservation/Rouge.lean, theorems overlap_le, rouge1_mem_unit, rouge1_self, rouge1_perm, at observation-data-mining f3914f0; what the check covers is stated in the book’s appendix C.

Used in

Data Mining as Observation chapters 0, 8.

Related

bag of words; perplexity; harness; certificate.

See also

Book equations stated beside the entry’s terms, not defining it: 0.22.

Ledger rows that cite the entry’s records without naming it: NEG-4.

Status

Generated 2026-09-10 by encyclopedia/generate.py; book at observation-data-mining f3914f0; the commit of every record is listed in the encyclopedia’s provenance.

← rotary position embeddingsafe pruning →