Hello Lou,

What I would like to link is:

1) "la" (no space after) in <orig> (I think this tag is more
appropriate here than <sic>) to "la" in <reg>

2) "cho sa" in <orig> to "cosa" in <reg>

I cannot figure out an easier way than something
like
<choice>
    <orig>
        <seg type="ot" rend="no_space_after">la</seg>
        <seg type="ot">
            <seg rend="space_after">cho</seg>
            <seg>sa</seg>
        </seg>
    </orig>
    <reg>
        <seg type="ot">la</seg>
        <seg type="ot">cosa</seg>
    </reg>
</choice>

and then use linking and alignment methods.

The question is:
If I want the tokenization be done automatically, how should
I pre-tag this kind of regulirazation?

It seems to me that Emmanuele's initial idea of introducing a new
tag makes sense.

I would suggest something like <sb/> for "space break" with
@type="orig|reg":

<choice>
    <orig>la<sb type="reg"/>cho<sb type="orig"/>sa</orig>
    <reg>la cosa</reg>
</choice>

or even

la <sb type="reg"/> <choice><orig>cho<sb type="orig"/>sa</orig>
<reg>cosa</reg></choice>

Best,

Ale(x|ks)e[iïyj]

----- Original Message ----- 
From: "Lou Burnard" <[log in to unmask]>
To: "Alexey LAVRENTEV" <[log in to unmask]>
Cc: <[log in to unmask]>
Sent: Monday, March 03, 2008 11:07 AM
Subject: Re: Critical Insert


> Hello Alexe[iy]!
>
> Thanks for the suggestion as to a possible use-case. But surely this could 
> be coped with quite easily:
>
> <choice>
> <sic><seg>lacho</seg><seg>sa</seg></sic>
> <reg><seg>la</seg> <seg>cosa</seg></reg>
> </choice>
>
> (I've used <seg> here rather than <w> to avoid arguments about what a word 
> is, but the same principles apply; you could also saying something like 
> <seg type="orthographicToken"> or introducing a new element <ot> for that 
> purpose!)
>
> Once you have the tokens identified, you can use the usual linking and 
> alignment methods to associate them, of course.
>
>
>
>
> Alexey LAVRENTEV wrote:
>> Lou Burnard wrote:
>>>
>>> What's wrong with
>>>
>>> <choice>
>>> <sic>lacho sa</sic>
>>> <reg>la cosa</reg>
>>> </choice>
>>>
>>> ?
>>>
>>> What kinds of query or analysis might you realistically do that this 
>>> encoding would not support?
>>
>> I am afraid that this markup makes a confusion between 2 different kinds
>> of regularization:
>> a) spelling,
>> b) spacing.
>>
>> In diplomatic editions, it is not uncommon to have normalized spacing
>> and original spelling.
>>
>> Such an encoding can also cause problems in case of linguistic 
>> tokenization.
>> If you tokenize on the basis of the regularized form but want to keep a 
>> track
>> of the original one, how will you link the two?
>>
>> Best,
>>
>> Alexei
>