BLEU: a Method for Automatic Evaluation of Machine Translation
Proceedings of the 40th Annual Meeting on Association for Computational Linguistics - ACL '02, 2001 · paper · Computational linguistics
Proposes BLEU, an automatic method for evaluating machine translation by comparing output with human reference translations through phrase matches. The authors present it as quick, inexpensive and language-independent, and as correlating highly with human evaluation.
Checked: PDF from aclanthology.org, fetched 5 Oct 2026. Page numbers are those of that PDF.
0 findings
None that passed a second check.
6 minor (typos and the like; most checked by the judge only)
Wording Text says "We computed the means, variances, and paired t-statistics which are displayed in Table 2" but Table 2 shows a "StdDev" row.
- p. 6 Section 3 text and Table 2
We computed the means, variances, and paired t-statistics which are displayed in Table 2.
Judge's note: Table 2 on page 6 lists Mean, StdDev, t; wording mismatch only.
Second check: another model agreed
Report a problem with this finding
Typo Stray space before comma: "fidelity , and fluency" (Section 1.1).
- p. 1 Section 1.1, first sentence
fidelity , and fluency
Judge's note: Crop shows a visible space between italic 'fidelity' and the comma.
Second check: none (judge only)
Report a problem with this finding
Typo Misspelling on p.4, Section 2.1.3: "a weighted average of the logarithm of modified precisions satisifies this requirement" should read "satisfies".
- p. 4 Section 2.1.3, end of first paragraph
satisifies this requirement
- p. 4
fied precisions satisifies this requirement
Judge's note: I checked the crop image of p.4, left column, at the end of Section 2.1.3. It clearly prints "satisifies", and the text layer (main.txt line 331) gives the same spelling, so this is not an extraction artefact. It is a cosmetic typo, and the fix is "satisfies".
Second check: none (judge only)
Report a problem with this finding
Typo Footnote 5 writes "N_max ≤ 4" while Section 2.3 denotes the maximum n-gram length as N ("we use N = 4").
- p. 4 Footnote 5
(for Nmax ≤4)
- p. 5 Section 2.3
In our baseline, we use N = 4
Judge's note: Both crops confirm the two notations; meaning clear, cosmetic inconsistency.
Second check: none (judge only)
Report a problem with this finding
Typo Figure 5/6 y-axes read "Monolingual Judgement"/"Bilingual Judgement" while text uses "judgment".
- p. 7 Figure 5 and Figure 6 y-axis labels
Monolingual Judgement
- p. 7 Section 5 body text
tracks human judgment well
Judge's note: Page 7 image shows 'Judgement' on axes; text and captions use 'Judgments'. Spelling variant inconsistency.
Second check: none (judge only)
Report a problem with this finding
Typo Acknowledgments: "reflect the position of policy of the Government" should read "position or policy".
- p. 8 Acknowledgments
reflect the position of pol- icy of the Government
Judge's note: Crop confirms 'position of pol-icy'; evident word slip in standard disclaimer.
Second check: none (judge only)