[ChineseGP] proposal to eliminate the divergence between us
HiroHOTTA
hotta at jprs.co.jp
Sun Jan 18 15:15:55 UTC 2015
Hi Chris,
JGP is investigating how each language GP should generate its LGR.
We finally have come to understand that, for example, JGP should
have at least "passively defined" variants that are defined in CGP
and/or KGP.
Borrowing Wang Wei's example,
Let's suppose
CGP defines
一4E00 (0); 一4E00(86),一4E00(886); 一(0),壱(0),壹(0),弌(0);
壱58F1 (0); 壹58F9(86),壹58F9(886); 一(0),壱(0),壹(0),弌(0);
壹58F9 (0); 壹58F9(86),壹58F9(886); 一(0),壱(0),壹(0),弌(0);
弌5F0C (0); 一4E00(86),一4E00(886); 一(0),壱(0),壹(0),弌(0);
JGP defines
'一' and '壱' as independent (non-variant) characters, and
'壹' and '弌' are not in JGP repertoire.
In such a case, JGP must define '一', '壱', '壹', and '弌' as
variants, and '一'and '壱' are both allocatable when '一'or '壱'
is applied for registration. In JGP, we call this as "passively
defined variants" - it seems that 'inherited' or 'adopted' is
used by Chris for this.
"Passively defined variants" is necessary if this is not
automatically calculated in an output of IP. I hope each language
GP and IP have a discussion on the possibility and specification
of this (automatic) calculation.
Hiro
On Fri, 9 Jan 2015 12:16:26 +0000
"Dillon, Chris" <c.dillon at ucl.ac.uk> wrote:
> Dear Qichao,
> Effectively the CN rule becomes a general rule once the mappings are made compatible. Variants in any language LGR may block labels even if they come from an LGR where they would not be blocked.
> I like that particular example especially, as it indicates a first come first served situation where a label including a low frequency character (and which is not a variant in all the language LGRs) blocks a label with a high frequency character.
> Again this is presuming the “each language has a descriptive LGR of its own situation (with or without variant mappings) and then another LGR (with possibly adopted* variant mappings) which is compatible with other LGRs for the script” model.
> *I have been using the word “inherited” for this, but “adopted” (i.e. from other languages using the script) may be better.
> Regards,
> Chris.
> --
> Research Associate in Linguistic Computing, Centre for Digital Humanities, UCL, Gower St, London WC1E 6BT Tel +44 20 7679 1599 (int 31599) www.ucl.ac.uk/dis/people/chrisdillon<http://www.ucl.ac.uk/dis/people/chrisdillon>
>
> From: ?超 [mailto:qichao at cnnic.cn]
> Sent: 09 January 2015 03:21
> To: Dillon, Chris; Sarmad Hussain
> Cc: Yoshiro YONEYA; ChineseGP at icann.org; hotta at jprs.co.jp
> Subject: Re: Re: [ChineseGP] proposal to eliminate the divergence between us
>
> Dear Chris,
>
> I agree with your understanding: if there is a variant mapping in any language, there will be the mapping in all language. So the mappings are formally unified, but C J and K still could apply different rule for allocated/blocked case.
>
> Your case is a nice example, but this is a minor error that the situation of
>
> 'where .XX壱 was allocated, disallowing the separate allocation of . XX一'
>
> is a rule only for language tag CN, while '.XX壱' and '.XX一 'could be both allocated for language tag JP or KR, and these are in a mapping similar to tag CN but different disposition rule.
>
> Maybe my understanding is wrong, and I hope IP can give an common example for C, J and K.
>
>
> Best Regards
>
> ________________________________
> ?超 via foxmail
>
> ?件人： Dillon, Chris<mailto:c.dillon at ucl.ac.uk>
> ?送??： 2015年1月8日(星期四) 下午9:04
> 收件人： Sarmad Hussain<mailto:sarmad.hussain at icann.org>
> 抄送： yoshiro.yoneya at jprs.co.jp<mailto:yoshiro.yoneya at jprs.co.jp>; ChineseGP at icann.org<mailto:ChineseGP at icann.org>; hotta at jprs.co.jp<mailto:hotta at jprs.co.jp>
> 主?： Re: [ChineseGP] proposal to eliminate the divergence between us
> Dear Sarmad,
> That is certainly an example of what I mean. The possible issue with CJK is that there will be many such examples. We will only know how many when we have tables for SC, TC, J and K.
> The CJK tables with compatible mappings* (i.e. not the tables which describe the situation in SC, TC, J and K) will probably end up with:
> “If it’s a variant in any language, it will be a variant in the tables with compatible mappings*.”
> *If my understanding is right, there will be no single amalgamated table.
> From a Japanese (Traditional Chinese and Korean) perspective, it Is weird to have, for example, 髮 ‘hair’ and 發 ‘send’ as variants (as they are both ? in Simplified Chinese) and on slide 8 of Professor Kim’s presentation, there was a possible situation where .XX壱 was allocated, disallowing the separate allocation of . XX一, although 一 is a much more basic character. However, I am yet to discover cases (and there would need to be quite a few of them) which could be party-stoppers.
> The Procedure is very well made and I believe it will be possible to follow it.
> Regards,
> Chris.
> --
> Research Associate in Linguistic Computing, Centre for Digital Humanities, UCL, Gower St, London WC1E 6BT Tel +44 20 7679 1599 (int 31599) www.ucl.ac.uk/dis/people/chrisdillon<http://www.ucl.ac.uk/dis/people/chrisdillon>
>
> From: Sarmad Hussain [mailto:sarmad.hussain at icann.org]
> Sent: 08 January 2015 11:01
> To: Dillon, Chris
> Cc: ChineseGP at icann.org<mailto:ChineseGP at icann.org>; Wang Wei; yoshiro.yoneya at jprs.co.jp<mailto:yoshiro.yoneya at jprs.co.jp>; hotta at jprs.co.jp<mailto:hotta at jprs.co.jp>
> Subject: RE: [ChineseGP] proposal to eliminate the divergence between us
> Dear Chris,
> If I understand this correctly, here is an example from Arabic script which could be relevant:
> U+06A9 (?) and U+06AA (?) are distinct letters (not variants) in Sindhi language (see http://www.omniglot.com/writing/sindhi.htm).
> However, these code points are considered variants of U+0643 (?) by other language communities (e.g. Arabic language).
> Therefore, they are being considered as variants by the ArabicGP (see Table 3 in https://community.icann.org/download/attachments/47253587/Arabic%20Variant%20Analysis%20for%20LGR%200.8.pdf?version=2&modificationDate=1419700233000&api=v2).
> Regards,
> Sarmad
> From: Dillon, Chris [mailto:c.dillon at ucl.ac.uk]
> Sent: Monday, January 05, 2015 3:09 PM
> To: Wang Wei; yoshiro.yoneya at jprs.co.jp<mailto:yoshiro.yoneya at jprs.co.jp>; hotta at jprs.co.jp<mailto:hotta at jprs.co.jp>
> Cc: ChineseGP at icann.org<mailto:ChineseGP at icann.org>; Sarmad Hussain
> Subject: RE: [ChineseGP] proposal to eliminate the divergence between us
> Dear colleagues,
> 新年快樂
> 明けましておめでとうございます
> ? ? 福 ?? ????
> Or/. Happy New Year!
> I am wondering whether there may be a way of making the proposal below work, without the JGP’s having to define variant sets and mappings (well, only a small number in scenario 2).
> Scripts used by many languages, for example Cyrillic and Arabic (I’ll leave out Latin as it is used by so many languages it may cause confusion) may be in a situation where some implementations of the script define variants (cf. SC and TC) and some don’t (cf. Japanese). One possible approach could be that languages which don’t define variants inherit the variant sets and mappings from the languages using the script that do define variants. I’m copying Sarmad in on this email, as this is a phenomenon which may have occurred in the work of one of the other GPs.
> I reckon this approach would work for cases 1, 3 and 4 below. (Actually 5 too as long as there are no examples of it…)
> That only leaves us with cases in scenario 2 such as 栞 (a variant which only exists in the Japanese table) for which a mapping to 刊 and 刋 would need to be created. For all other cases, the SC/TC mappings would be inherited.
> Regards,
> Chris.
> --
> Research Associate in Linguistic Computing, Centre for Digital Humanities, UCL, Gower St, London WC1E 6BT Tel +44 20 7679 1599 (int 31599) www.ucl.ac.uk/dis/people/chrisdillon<http://www.ucl.ac.uk/dis/people/chrisdillon>
>
> From: chinesegp-bounces at icann.org<mailto:chinesegp-bounces at icann.org> [mailto:chinesegp-bounces at icann.org] On Behalf Of Wang Wei
> Sent: 29 December 2014 07:54
> To: yoshiro.yoneya at jprs.co.jp<mailto:yoshiro.yoneya at jprs.co.jp>; hotta at jprs.co.jp<mailto:hotta at jprs.co.jp>
> Cc: ChineseGP at icann.org<mailto:ChineseGP at icann.org>
> Subject: [ChineseGP] proposal to eliminate the divergence between us
>
> Dear Yoneya San and Hotta San
>
> Please kindly accept my belated but best wishes for the Christmas and new year.
> Recently, we carried out the following works and I outlined them here for your comments:
>
> For any Hanzi in CGP repertoire, it belong to a variant mapping set (minimum set size is 1 which means there is no variant for the code point) under the current rules borrowed from CDNC; and for any Kanji code point in JGP repertoire, it may also belong to some variant mapping set (we acknowledge that there is no variant in JPRS practice so far, but we assume that there will be a kind of variant mapping definition in JGP repertoire).
>
> All the variant mapping sets can be divided into FIVE scenarios:
>
>
>
> 1. the variant mapping set in JPRS ∈ variant mapping set in CDNC
> [cid:image001.jpg at 01D02C04.FD28B860]
> In CGP
> 愛 611B (0);?(86),愛(886);愛(0),?(0);
> ? 7231 (0);?(86),愛(886);愛(0),?(0);
>
> In JGP
> 愛611B(2,3);611B(2,3);
>
>
>
>
> 2. the variant mapping set in CDNC ∈ the variant mapping set in JPRS
> [cid:image002.jpg at 01D02C04.FD28B860]
> In CGP:
> 刊520A (0);刊520A(86),刊520A(886);刊(0),刋(0);
> 刋520B (0);刊520A(86),刊520A(886);刊(0),刋(0);
>
> In JGP:
> 刊 520A(2,3);520A(2,3);
> 刋 520B(2,3);520B(2,3);
> 栞 681E(2,3);681E(2,3);
> *: this example is ONLY an assumption
>
>
>
>
> 3. the variant mapping set in CDNC = the variant mapping set in JPRS
> [cid:image003.jpg at 01D02C04.FD28B860]
> In CGP
> 一4E00 (0); 一4E00(86),一4E00(886); 一(0),壱(0),壹(0),弌(0);
> 壱58F1 (0); 壹58F9(86),壹58F9(886); 一(0),壱(0),壹(0),弌(0);
> 壹58F9 (0); 壹58F9(86),壹58F9(886); 一(0),壱(0),壹(0),弌(0);
> 弌5F0C (0); 一4E00(86),一4E00(886); 一(0),壱(0),壹(0),弌(0);
>
> In JGP:
> 一 4E00(2,3);4E00(2,3);
> 壱 58F1(2,3);58F1(2,3);
> 壹 58F9(2,3);58F9(2,3);
> 弌 5F0C(2,3);5F0C(2,3);
> *: this example is ONLY an assumption
>
>
>
>
> 4. the variant mapping set in CDNC ∩ the variant mapping set in JPRS = 0
> [cid:image004.jpg at 01D02C04.FD28B860]
> The code point UNIQUELY exists in JGP table
> 辻8FBB(2,3);8FBB(2,3);
>
>
>
>
> 5. the variant mapping set in CDNC ∩ the variant mapping set in JPRS ≠ 0
>
> and
> the variant mapping set in CDNC ≠ the variant mapping set in JPRS
> [cid:image005.jpg at 01D02C04.FD28B860]
>
> No specified example so far
>
> \
>
> In the past, we discussed the variants problem for many times, but mainly based on the two types: allocatable and blocked. However, we think another type ("out-of- repertoire") in the XML draft, may help the conflicted issue between JGP and CGP, which was recommended by Asmus' mail.
> The basic principle is "any variant label with a code point out-of-repertoire is invalid". We think this “out-of-repertoire” type and consequent “invalid” action will tremendously decrease the complexity of variant mapping coordination between us.
>
> For scenario 1:
> In CGP
> 愛 611B (0);?(86),愛(886);愛(0),?(0);
> ? 7231 (0);?(86),愛(886);愛(0),?(0);
> In JGP
> 愛 611B(2,3);611B(2,3);
>
> JGP take? 7231 into variant mapping set, but mark it as “out-of-repertoire” and take “invalid” action for WLG process, which means, ? 7231 will never be generated into the labels.
>
> JGP LGR:
> <language>und-Jpan</language>
> <char cp="611B" tag="sc:Hani">
> <var cp="611B" type="alloc" comment="identity" />
> <var cp="7231" type="out-of-repertoire-var" /> <!--Hans, JGP should exist.-->
> </char>
> WLE rules:
> <action disposition="invalid" any-variant="out-of-repertoire-var"
> comment="any variant label with a code point out of repertoire is invalid"/>
> <action disp="allocatable" all-variant="alloc" />
>
> CGP LGR:
> <language>und-Hani</language>
> <char cp="611B" tag="sc:Hani">
> <var cp="611B" type="trad" comment="identity" /> <!-- Jpan -->
> <var cp="7231" type="simp" />
> </char>
> <char cp="7231" tag="sc:Hani">
> <var cp="611B" type="trad" /> <!-- Jpan -->
> <var cp="7231" type="simp" comment="identity" />
> </char>
> WLE rules:
> <action disp="blocked" any-variant="block" />
> <action disp="allocatable" only-variants="simp both" />
> <action disp="allocatable" only-variants="trad both" />
> <action disp="blocked" any-variant="simp trad" />
> <action disp="allocatable" comment="catch-all" />
>
>
> For scenario 2:
> In CGP:
> 刊520A (0);刊520A(86),刊520A(886);刊(0),刋(0);
> 刋520B (0);刊520A(86),刊520A(886);刊(0),刋(0);
>
> In JGP:
> 刊 520A(2,3);520A(2,3);
> 刋 520B(2,3);520B(2,3);
> 栞 681E(2,3);681E(2,3);
>
> Now it is CGP’s turn to take栞 681E into variant mapping set, but mark it as “out-of-repertoire” and take “invalid” action for WLG process, which means, 栞 681E will never be generated into the labels.
>
> CGP LGR
> <language>und-Hani</language>
> <char cp="520A" tag="sc:Hani">
> <var cp="520A" type="both" comment="identity" />
> <var cp="520B" type="block" />
> <var cp="681E" type="out-of-repertoire-var" /> <!-- Jpan -->
> </char>
> <char cp="520B" tag="sc:Hani">
> <var cp="520A" type="both" />
> <var cp="520B" type="block" comment="identity" />
> <var cp="681E" type="out-of-repertoire-var" /> <!-- Jpan -->
> </char>
> <char cp="681E" tag="sc:Hani"> <!-- Jpan -->
> <var cp="520A" type="block" />
> <var cp="520B" type="block" />
> <var cp="681E" type="out-of-repertoire-var" comment="identity"/>
> </char>
> WLE rules:
> <action disp="invalid" any-variant="out-of-repertoire-var"
> comment="any variant label with a code point out of repertoire is invalid"/>
> <action disp="blocked" any-variant="block" />
> <action disp="allocatable" only-variants="simp both" />
> <action disp="allocatable" only-variants="trad both" />
> <action disp="blocked" any-variant="simp trad" />
> <action disp="allocatable" comment="catch-all" />
>
> JGP LGR:
> <language>und-Jpan</language>
> <char cp="520A" tag="sc:Hani">
> <var cp="520A" type="alloc" comment="identity" />
> <var cp="520B" type="block" />
> <var cp="681E" type="block" />
> </char>
> <char cp="520B" tag="sc:Hani">
> <var cp="520A" type="block" />
> <var cp="520B" type="alloc" comment="identity" />
> <var cp="681E" type="block" />
> </char>
> <char cp="681E" tag="sc:Hani">
> <var cp="520A" type="block" />
> <var cp="520B" type="block" />
> <var cp="681E" type="alloc" comment="identity"/>
> </char>
> WLE rules:
> <action disp="blocked" any-variant="block" />
> <action disp="allocatable" all-variant="alloc" />
>
>
>
> For Scenario 3:
>
> In CGP
> 一4E00 (0); 一4E00(86),一4E00(886); 一(0),壱(0),壹(0),弌(0);
> 壱58F1 (0); 壹58F9(86),壹58F9(886); 一(0),壱(0),壹(0),弌(0);
> 壹58F9 (0); 壹58F9(86),壹58F9(886); 一(0),壱(0),壹(0),弌(0);
> 弌5F0C (0); 一4E00(86),一4E00(886); 一(0),壱(0),壹(0),弌(0);
>
> In JGP:
> 一 4E00(2,3);4E00(2,3);
> 壱 58F1(2,3);58F1(2,3);
> 壹 58F9(2,3);58F9(2,3);
> 弌 5F0C(2,3);5F0C(2,3);
>
> JGP needs to create its own mapping set including all above 4 code points and corresponding rules, otherwise, it will fall into scenario 1..
>
>
> For Scenario 4:
> Like UNIQUE code point ONLY exists in JGP table
> 辻8FBB(2,3);8FBB(2,3);
>
>
> CGP probably will not include this code point into its repertoire.
> No extra work or rule are needed.
>
>
> For Scenario 5:
> [cid:image006.jpg at 01D02C04.FD28B860]
>
> Actually, we have not find the code points which fit into this scenario.
> But the solution will refer to scenario 1 or 2, like:
>
> For JGP, “C” will be included but marked as “out-of-repertoire”
> For CGP, “A” will be included but marked as “out-of-repertoire”
>
>
> In conclusion, “out-of ?repertoire type” and “invalid action” provide us a conservative and simple way to reach a consensus for the variant mapping and rules.
> According to our analysis on CGP table and JPRS table
> There are 4983 code points fit for Scenario 1
> There are 840 code points fit for Scenario 3
> There are 170 code points fit for Scenario 4
>
> Since JGP has not decided yet if variant relationship exist in JGP repertoire, we don’t have analytical number about scenario 3 and scenario 5. But what we believe is that the above solution can also be applied for scenario 3 and 5 no matter what kind of variant mapping JGP will produce.
>
>
> All above is our proposal for settle the divergence at minimum cost for both of us.
> What do you think about it? Looking forward for your reply.
>
>
> Best Regards,
> Wei Wang
More information about the ChineseGP
mailing list