Treatment of adverbial "at all"

[manning]

How should we treat adverbial "at all"? At the moment the dependencies treat "at" as case on all, but then surely its POS tag should no longer be ADV?

8	not	not	ADV	RB	_	9	advmod	_	_
9	going	go	VERB	VBG	VerbForm=Ger	6	advcl	_	_
10	at	at	ADV	RB	_	11	case	_	_
11	all	all	ADV	RB	_	9	obl	_	SpaceAfter=No

9	at	at	ADV	RB	_	10	case	_	_
10	all	all	ADV	RB	_	6	obl	_	SpaceAfter=No

7	does	do	AUX	VBZ	Mood=Ind|Number=Sing|Person=3|Tense=Pres|VerbForm=Fin	9	aux	_	_
8	not	not	PART	RB	_	9	advmod	_	_
9	help	help	VERB	VB	VerbForm=Inf	3	parataxis	_	_
10	me	I	PRON	PRP	Case=Acc|Number=Sing|Person=1|PronType=Prs	9	obj	_	_
11	at	at	ADV	RB	_	12	case	_	_
12	all	all	ADV	RB	_	9	obl	_	_

[jnivre]

Isn't "at all" a good candidate for a fixed expression? The fact that both "at" an "all" have the tag ADV seem to indicate that this is what the annotator thought.

[dan-zeman]

+1 on fixed. But then there is no room for the ADV tag, except for the MISC column (I think we have been using it to store something like MWE_POS, if I recall the name of the attribute correctly.)

[arademaker]

We are using MWE_POS in the MISC column in the UD_Portuguese corpus

[sebschu]

This should already be fixed according to our guidelines (http://universaldependencies.org/en/dep/mwe.html).

[jnivre]

Actually not. The page you refer to has a list of things annotated as mwe/fixed, but it also has a list of things NOT annotated in this way. And "at all" is on the latter.

[sebschu]

Yes, of course. Sorry for the confusion!

[nschneid]

Now at/ADP all/DET in both EWT and GUM