Each as PRON instead of DET

[AngledLuffa]

The standard in English treebanks is to label each as a DET:

# sent_id = weblog-blogspot.com_alaindewitt_20040929103700_ENG_20040929_103700-0079
# text = His column appears in The Hill each week.
1       His     his     PRON    PRP$    Case=Gen|Gender=Masc|Number=Sing|Person=3|Poss=Yes|PronType=Prs 2       nmod:poss       2:nmod:poss     _
2       column  column  NOUN    NN      Number=Sing     3       nsubj   3:nsubj _
3       appears appear  VERB    VBZ     Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin   0       root    0:root  _
4       in      in      ADP     IN      _       6       case    6:case  _
5       The     the     DET     DT      Definite=Def|PronType=Art       6       det     6:det   _
6       Hill    Hill    PROPN   NNP     Number=Sing     3       obl     3:obl:in        _
7       each    each    DET     DT      _       8       det     8:det   _
8       week    week    NOUN    NN      Number=Sing     3       obl:tmod        3:obl:tmod      SpaceAfter=No
9       .       .       PUNCT   .       _       3       punct   3:punct _

Any objections to updating the each UPOS from PRON to DET?

[rueter]

I trust this is on a by-context basis. I can also think of an ADV reading:

They gave us two apples each.

[AngledLuffa]

I will do it as a PR before making any final changes

…

On Fri, Oct 27, 2023, 9:56 AM Jack Rueter ***@***.***> wrote: I trust this is on a by-context basis. I can also think of an ADV reading: *They gave us two apples each.* — Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA2AYWK4Y3QP2S5TOXV6633YBPRTNAVCNFSM6AAAAAA6TC6YTWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTGIZDONBUGQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[nschneid]

I think EWT probably has a consistent policy on "each". (Demonstratives not in det function become PRON, but most other DETs stay DET: https://universaldependencies.org/en/pos/DET.html)

[AngledLuffa]

@nschneid that's not what happens in practice in EWT. each other gets tagged DET with the feature ExtPos=PRON, and all of the others just get tagged DET. For example:

# the spaces are each are actually tabs
grep " each    " extern_data/ud2/git/UD_English-EWT/en_ewt-ud-train.conllu | cut -f 4 | sort | uniq
DET

[nschneid]

Right, I'm saying only demonstratives and relativizers become PRON in EWT. "each" is neither so it stays DET.

[AngledLuffa]

Got it. In that case, unless there are objections from the stakeholders, I'll set each each to DET

…

On Fri, Oct 27, 2023, 1:36 PM Nathan Schneider ***@***.***> wrote: Right, I'm saying only demonstratives and relativizers become PRON in EWT. "each" is neither so it stays DET. — Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA2AYWMZM5GAK5K5POLKWPTYBQLNBAVCNFSM6AAAAAA6TC6YTWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTGQ4DIMJRG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[LarsAhrenberg]

In that case, unless there are objections from the stakeholders,

I do object to anyone making changes to UD_English-LinES without my consent. I welcome that errors are pointed out, where they exist, but I much prefer to correct them myself. So, please don't.

While I appreciate the efforts to improve consistency between treebanks I fear it sometimes is going too fast. Many decisions are now being made relating to EWT and GUM that revoke earlier decisions and where's the guarantee that they will not be changed in the future? And minor inconsistencies cannot be that harmful. Moreover, increasing consistency in one place may cause inconsistency in others, e.g. between languages. I have annotated data that I maintain to add to the treebank in the future and that will obviously be less consistent with the public UD data if anyone can just go in and make changes. Besides, I think the decision to declare each as always DET is misguided and one that may well be changed in the future. See the example above from @rueter.

[AngledLuffa]

That's fair, and it's not my intention to step on toes. If you want the "each" change unrolled, I can certainly do that. I do think the intention is to have UPOS be truly universal, though, and the expectation would be for the same sentence to have the same tags across treebanks. If there are disagreements between treebank annotation schemes, it would be good to come to a consensus.

…

On Sat, Oct 28, 2023, 12:36 AM Lars Ahrenberg ***@***.***> wrote: In that case, unless there are objections from the stakeholders, I do object to anyone making changes to UD_English-LinES without my consent. I welcome that errors are pointed out, where they exist, but I much prefer to correct them myself. So, please don't. While I appreciate the efforts to improve consistency between treebanks I fear it sometimes is going too fast. Many decisions are now being made relating to EWT and GUM that revoke earlier decisions and where's the guarantee that they will not be changed in the future? And minor inconsistencies cannot be that harmful. Moreover, increasing consistency in one place may cause inconsistency in others, e.g. between languages. I have annotated data that I maintain to add to the treebank in the future and that will obviously be less consistent with the public UD data if anyone can just go in and make changes. Besides, I think the decision to declare *each* as always DET is misguided and one that may well be changed in the future. See the example above from @rueter <https://github.com/rueter>. — Reply to this email directly, view it on GitHub <#11 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA2AYWJTQY3ZEHE76CL4AMLYBSYYTAVCNFSM6AAAAAA6TC6YTWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTOOBTG4ZTKMJQHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[LarsAhrenberg]

To be consistent with myself I'll do a review of each before Nov. 1st and change back, where necessary.