Jump to content

Talk:BLEU

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
Good articleBLEU has been listed as one of the Language and literature good articles under the good article criteria. If you can improve it further, please do so. If it no longer meets these criteria, you can reassess it.
Article milestones
DateProcessResult
March 11, 2007Good article nomineeNot listed
March 29, 2007Good article nomineeListed
March 11, 2009Good article reassessmentKept
Current status: Good article

oops missing info?

[edit]

"This is illustrated in the following example from," from where? --Ling.Nut 16:45, 13 March 2007 (UTC)[reply]

Well spotted. This example is from Papineni 2002. I've noted this. - Francis Tyers · 16:47, 13 March 2007 (UTC)[reply]

No pronoun antecedent

[edit]

There are a number of 3rd person plural pronouns in the Performance section which have no obvious antecedent--e.g. "they highlight". Perhaps a sentence has been left out referring to a study by some researchers? Or is it one of the footnoted references, and if so, which one? Mcswell (talk) 23:13, 29 November 2007 (UTC)[reply]

GA/R ended

[edit]

In accordance with the WP:GA/R over this article, in a 3 to 1 vote, it has been listed as a WP:GA. Review archived in GA/R archive 15. Homestarmy 15:55, 2 April 2007 (UTC)[reply]

algorithm candidate phrase?

[edit]

is this correct - the candidate phrase is supposed to be the the the the the the the?

Example of poor machine translation output with high precision
Candidate the the the the the the the
Reference 1 the cat is on the mat
Reference 2 there is a cat on the mat

it doesn't seem to be from the text, but it's been that way since the algorithm section was first inserted. --Ludwigs2 21:33, 29 March 2009 (UTC)[reply]

It is correct. See 'Example 2' in the BLEU paper. - Francis Tyers · 11:28, 19 June 2009 (UTC)[reply]

Interactive Scoring Tool

[edit]

I am the author of a freely available interactive BLEU scoring tool that is available at http://ibleu.googlecode.com. I believe that this information should be worth including in this BLEU wikipedia article. However, i wanted to run this by folks to make sure it was okay. There is also a peer-reviewed publication for the tool that can be cited [1].

[1] BLEU: Interactively Scoring and Debugging Statistical Machine Translation Systems. Proceedings of the Fifth IEEE International Conference on Semantic Computing (Demos). Nitin Madnani.

The publication is going to be published in the proceedings of an upcoming conference (in September 2011) as can be seen at http://ieee-icsc.org. — Preceding unsigned comment added by DesiLinguist (talkcontribs) 17:43, 23 July 2011 (UTC)[reply]

No longer true fact of BLEU

[edit]

This is sentence in the article "BLEU was one of the first metrics to achieve a high correlation with human judgements of quality," is no longer true. Papers dated back from 2006 already warns of BLEU and urges researchers to re-evaluate BLEU (reference: http://www.aclweb.org/anthology/E06-1032). --Alvations (talk) 00:54, 12 January 2017‎

Hi Alvas. I find it weird that an expression "X was the first Y" can be no longer true. Do you mean that, in that moment, they were already other metrics? High correlation is, of course, not a safe/clear claim, but in that moment the authors showed the correlation. "The high correlation coefficient of 0.99 indicates that BLEU tracks human judgment well" (Papineni et al 2002). I'll rephrased saying "claim" instead of achieve. Best. --Hectorpal (talk) 13:20, 18 October 2017 (UTC)[reply]

endnote [12] glitch

[edit]

The endnote for this sentence is rendered "[12]" (in Chrome Version 87.0.4280)

"BLEU cannot, in its present form, deal with languages lacking word boundaries.[12]"

However, the source code indicates the endnote is https://en.wikipedia.org/wiki/BLEU#endnote_Denoual2005a

It appears that the endnote should be rendered "[9]" instead of "[12]"

Specifically [9] Denoual, E. and Lepage, Y. (2005) Denoual, E. and Lepage, Y. (2005) "BLEU in characters: towards automatic MT evaluation in languages without word delimiters" in Companion Volume to the Proceedings of the Second International Joint Conference on Natural Language Processing pp. 81–86 — Preceding unsigned comment added by Gene.S.Miller (talkcontribs) 22:30, 15 January 2021 (UTC)[reply]

Page is broken and has been tampered with

[edit]

I tried to report a bug with Wikimedia, but their bug reporting tool was itself broken and very problematic!

The citation links on this page, for example things like "[7]", don't work properly when you click on them: they don't open up the relevant "page footer" that displays the citation. This is a bug because Wikipedia software should be checking and validating the syntax and functionality of the page before modifying it.

Furthermore, when you look at the links below, such as in the bibliography section, vandals have linked the citations or bibliography to spam sites. I'm not sure how these vandals are normally tracked down using their IP address or whatever and what should be done about them. 2001:569:76C2:4900:4CD8:99FF:FE05:A2A9 (talk) 01:35, 6 May 2022 (UTC)[reply]

More citation/endnote issues

[edit]

Could a more active editor investigate this article's citations? The citation system currently used can't be right -- there are both "Notes" and "References.". Like User:Gene.S.Miller above, (sorry, don't know how to @ mention on this platform, even after all these years), I found this very confusing.

Part of the issue is that both note and reference links look exactly the same in the text -- so Note [1] and Reference [1] cannot be distinguished without following them.

What are we to do? I can't remember a consensus on one style over the other, although I see more and more pages using the <ref> style.--Yoderj (talk) 16:50, 12 February 2024 (UTC)[reply]

Geometric mean

[edit]

Why is there no root for the geometric mean? 108.53.169.6 (talk) 02:01, 26 July 2024 (UTC)[reply]