W3C home > Mailing lists > Public > public-multilingualweb-lt@w3.org > May 2013

RE: [All] mostly minor issues on its2 draft and a question

From: Yves Savourel <ysavourel@enlaso.com>
Date: Tue, 7 May 2013 08:33:06 +0200
To: "'Felix Sasaki'" <fsasaki@w3.org>, 'Mārcis Pinnis' <marcis.pinnis@Tilde.lv>
CC: <public-multilingualweb-lt@w3.org>
Message-ID: <000801ce4aec$bd8c9f50$38a5ddf0$@com>
Hum... one more statement to check.

Besides there could be a use case where a tool auto detect term candidates and a human checks the result with term='yes|no' and a confidence score: "I'm pretty sure this should not go into our glossary, but someone needs to make sure of that."

I guess: why restrict the possibilities of what one can do if there is no harm (and it gives us no extra work)?

And, after all, logically <span its:term='yes' its:termConfidence='0.6'>Bled</span> is the same as <span its:term='no' its:termConfidence='0.4'>Bled</span> :)

-ys

-----Original Message-----
From: Felix Sasaki [mailto:fsasaki@w3.org] 
Sent: Tuesday, May 07, 2013 8:20 AM
To: Mārcis Pinnis
Cc: Yves Savourel; public-multilingualweb-lt@w3.org
Subject: Re: [All] mostly minor issues on its2 draft and a question

Am 07.05.13 07:57, schrieb Mārcis Pinnis:
> Hi Yves,
>
> I agree. If we do not want to prohibit the termConfidence usage also with term="no", then it would be better to remove the term="yes" part in the spec. However, as I said, I have not seen any tool that would need to explicitly add term="no" mark-up. Therefore, I also believe that the current version, although being restrictive, would not cause real restrictions. But ... if we want (have) to be correct then the "term="yes"" part should probably be removed.

Couldn't we also just say "term=no MUST NOT be used with termConfidence"? It sounds easier than removing term=yes ....

Best,

Felix

>
> Best regards,
> Mārcis ;o)
>
> -----Original Message-----
> From: Yves Savourel [mailto:ysavourel@enlaso.com]
> Sent: Tuesday, May 07, 2013 1:01 AM
> To: Mārcis Pinnis; 'Felix Sasaki'
> Cc: public-multilingualweb-lt@w3.org
> Subject: RE: [All] mostly minor issues on its2 draft and a question
>
> Hi Mārcis,
>
> So from an implementation viewpoint, do we want to have some kind of restriction with term='no' and termCofidence?
> I'm guessing it's probably simpler to just allow it with any valid value.
> But then maybe the text need to be changed and not mentioned explicitely term='yes'.
>
> cheers,
> -ys
>
> -----Original Message-----
> From: Mārcis Pinnis [mailto:marcis.pinnis@Tilde.lv]
> Sent: Monday, May 06, 2013 8:34 PM
> To: Felix Sasaki; Yves Savourel
> Cc: public-multilingualweb-lt@w3.org
> Subject: RE: [All] mostly minor issues on its2 draft and a question
>
> Hi Felix, Yves, All,
>
> Automated term extraction tools (as far as I have seen) do not mark phrases as non-terms. Practically, term extraction tools specify that a phrase IS A TERM with a certain confidence. Just theoretically, I can imagine that there might be a tool that performs also non-term annotations, but ... it is very unlikely. The term="no" as I see it is a means for human experts to explicitly say that a phrase is definitely not a term for some very specific cases that some term extraction tools might get wrong or tend to get wrong. But that means that the confidence in such a scenario is 1.0.
>
> I hope this helps further ...
>
> Best regards,
> Mārcis ;o)
>
> -----Original Message-----
> From: Felix Sasaki [mailto:fsasaki@w3.org]
> Sent: Sunday, May 5, 2013 6:04 PM
> To: Yves Savourel
> Cc: public-multilingualweb-lt@w3.org; Mārcis Pinnis
> Subject: Re: [All] mostly minor issues on its2 draft and a question
>
> Hi Yves, with CC to Mārcis,
>
> Am 04.05.13 16:03, schrieb Yves Savourel:
>> Hi Felix,
>>
>> sorry for the late answer:
>>
>>> "termConfidence represents the confidence of the agents producing 
>>> the annotation that the the annotation is a term (that is, term="yes").
>>> It does provide confidence incormation related to termInfoRef."
>> So what does termConfidence means when we have term='no'?
>> Doesn't it express the degree of confidence that the span of content is not a term?
> Good questions. I made the edits, see
> http://www.w3.org/International/multilingualweb/lt/drafts/its20/its20.
> html#terminology-local
> but maybe Mārcis can clarifiy this?
>
> Best,
>
> Felix
>
>> cheers,
>> -ys
>>
>>
>> -----Original Message-----
>> From: Felix Sasaki [mailto:fsasaki@w3.org]
>> Sent: Thursday, May 02, 2013 11:14 AM
>> To: public-multilingualweb-lt@w3.org
>> Subject: [All] mostly minor issues on its2 draft and a question
>>
>> Hi all,
>>
>> here are some mini issues I found in the ITS2 draft.
>>
>> - Sec. 4.3: Conformance Type 3: Processing Expectations for ITS Markup in HTML "If an application claims to process ITS markup implementing the conformance clauses 3-1, 3-2 and 3-3": delete "and 3-3", since this sentence is part of clause 3-3.
>>
>>
>> - Terminology sec. 8.4.2 "local markup" section says "An optional 
>> termConfidence attribute with the value of a rational number in the 
>> interval 0 to 1 (inclusive). The value follows the XML Schema double 
>> data type with the constraining facets minInclusive set to
>> 0 and maxInclusive set to 1. termConfidence represents the confidence of the agents producing the annotation that the values of the term and, where provided, termInfoRef, are accurate."
>> This is wrong: Marcis said that "termConfidence" relates only to the question whether the selected item is a node or not. This is also written in a note in the "text analyis" section:
>> "The confidence value applies to two pieces of information (see following rows in this table). This is opposed to termConfidence which is part of the Terminology data category. termConfidence represents the confidence in just a single piece of information: the decision whether something is a term or not (term). termConfidence does not relate to the confidence about additional information about the term that can be encoded with termInfoRef."
>>
>> So we can rewrite in sec. 8.4.2 instead of "termConfidence represents the confidence of the agents producing the annotation that the values of the term and, where provided, termInfoRef, are accurate."
>> the following
>> "termConfidence represents the confidence of the agents producing the annotation that the the annotation is a term (that is, term="yes"). It does provide confidence incormation related to termInfoRef."
>>
>>
>> If nobody disagrees I'll make the edits Friday night (European time).
>>
>>
>> Best,
>>
>> Felix
>>
>>
>
Received on Tuesday, 7 May 2013 06:33:38 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:32:08 UTC