search

LingSIG / wordAttributes

work space for a coherent proposal for inline attributes of <w> in TEI XML
1 stars 1 forks source link

Initial comments #3

Closed lb42 closed 6 years ago

lb42 commented 7 years ago
  1. @pos and @feats overlap considerably with the functionality of existing @ana : you need to explain what the differences are and why one might prefer one or the other. I also wonder whether anyone would ever really want to use them simultaneously.
  2. @reg still seems unnecessary to me : what's wrong with doing things properly ?
  3. @join seems like a cool idea though, probably because we already have something very like it on pc/@force : I'm surprised no-one is proposing a "multiword" element instead though.
  4. @lemma is, as you say, already there. So why is it on this list?
bansp commented 7 years ago

Thanks, Lou. 1) it's fairly standard to use @pos and @feats simultaneously, unless combined information is used, usually constructed in a positional fashion, with pos coming first and followed by a list of grammatical features. Still, not a bad point for discussion, given that we should minimize the impact on the existing TEI model, to maximize the chance of getting the proposal accepted. 2) "properly" has a weasel way of meaning different things to different people; in this case @reg is a well-attested compromise of providing the regularized value as an attribute, without exploding the markup. 3) right, we discussed pc/@pre briefly (or did you really mean @force?), as a rather limited tool that @join generalizes over (in terms of both the domain and the range: it would apply to all token-level elements and would be capable of expressing more than two values) 4) oh, I believe just for the sake of completeness; maybe I failed to clearly distinguish between the complete target vision and what exactly could/should be added to the already existing repertoire -- will have a look, thanks

lb42 commented 7 years ago
  1. "Fairly standard" meaning "fairly common", I assume (considering your comments elsewhere on the use of the word "standard" :-)
  2. "well-attested" -- any evidence for that? I have lots of evidence of people happily using <choice> to handle regularisation in historical corpora btw. And why adding a gazillion attributes doesn't count as "exploding the markup" you can explain to me another day.
  3. No, I really meant @force. But maybe I misunderstand what @join is for. Will check.
bansp commented 7 years ago
  1. My comments elsewhere concerned the noun "standard" in the phrase "developing a standard" :-) But I take your point -- I have used weasel statements myself here, indeed...
  2. ... which also includes the "well-attested". Maybe indeed the reductionist exercise here should yield a proposal for a single @pos attribute that could potentially be internally complex (and concatenate the information on the pos and on the morphosyntactic features).
  3. I'm a bit lost and will study @force to see how it could be used.
  4. Tomaz in issue #3 provides an argument against using @reg (which corroborates your point 2. in the initial note).
bansp commented 6 years ago

Thanks for your remarks, Lou -- they have improved the eventual proposal. Closing this issue after the merger with TEI/dev .