attributes of <w> (was: venting on another topic)

Hi Lou and All,

The issue has been raised several times in the context of LingSIG. These 
are all fine attributes, but referencing other places in or out of the 
document is sometimes a technical problem in mass processing of data, as 
in huge corpora (and we have some such). Being able to place limited 
info inside <w> locally would sometimes be hugely helpful (read also: it 
would hopefully not make corpus analysts turn to other encoding formats, 
as they naughtily do, all to often).

Limited as in, e.g. @pos for POS, @msd for morphosyntactic description. 
Together with @lemma, the three would allow us to conquer half the 
world... I promise.

:-)

On 08/10/16 18:23, Lou Burnard wrote:
> On 07/10/16 21:11, Martin Mueller wrote:
>>
>> David Sewell’s post makes me utter another wish: why not allow
>>  attributes  to <w> elements that let you introduce light linguistic
>> annotation into a text in a manner that does not interfere with the
>> larger structure of the document and is easier to process.
>>
>
> You mean  the existing @type, @subtype, @ana, @function,  @datcat,
> @valueDataCat, @lemma, @lemmaRef   (etc)  attributes are insufficient?
>
>>
>>
>> I’m thinking specifically of  ‘pos’ and ‘reg’ as annotations that one
>> might want to add to a word token in addition to ‘lemma’.
>>
>
> If "pos" means "part of speech", what's wrong with @ana, @function, or
> @datcat ?
>
>> It has never been quite clear to me why the “War on Attributes” some
>> years ago abolished attributes of this kind.
>>
>
> It didn't. It abolished the attribute "@reg" on the grounds that you
> might well want to include in its value some markup (e.g. a <g> element)
> which would then not be correctly processed since attribute values may
> not contain markup. But in the simple case you could use @lemma to hold
> a regularized form I suppose.
>
>>
>> The Stuttgart corpus query language had the very useful concept of
>> “positional attributes”. The BlackLab uses the same concept, and it is
>> very intuitive to think of  a word token as something that has a
>> number of “properties” among which the spelling at a location is
>> privileged in some ways.  If you allow for the annotation of a <w>
>> with attributes, you create an intuitively plausible environment in
>> which the actual textual data appear in the text stream, but
>> linguistic metadata are added as properties of the token.
>>
>
> Yes, that's exactly what we do.
>
>> whether other people have had similar thoughts.
>
>

-- 
Piotr Bański, Ph.D.
Senior Researcher,
Institut für Deutsche Sprache,
R5 6-13
68-161 Mannheim, Germany