1 Comments to: Referential Structures and Links (S. Salmon-Alt and L. Romary) Massimo Poesio University of Essex, Language & Computation Group.

1

Comments to:Referential Structures and Links(S. Salmon-Alt and L. Romary)

Massimo PoesioUniversity of Essex, Language & Computation Group

ISO/SIGSEM Workshop on Multim. Sem. Repr - Tilburg, Jan 20052

Summary of the RAF proposal

• A Reference Annotation Framework (RAF) instantiating Ide & Romary’s LAF

• TWO main `data categories’ (= elements):– MARKABLES– LINKS

• Recommendations about descriptors of markables and links


Overview of comments

• I share much of the philosophy behind the proposal – Markables – Separate links– Useful attributes for reference

• But (some of these issues already covered in presentation)– Some aspects may be more ready for standardization than others– Are all the proposed descriptors really part of ‘reference’

annotation?– Practical considerations to think about:

• Use• Annotation technology (where does standoff fit in?)

• Some difficult issues still to be addressed


Separate links

• Case for links as independent elements: a form of normalization– (Ambiguity – see below)

– Multiple relations (identity, bridging, more general ACE-style semantic relations)

– Attributes of relations (type of relation, different discourse models)

• BUT: MMAX-style set-based annotation more transparent representation of discourse models


Multiple relations in the GNOME corpus

Cabinet on Stand

The decoration on this monumental cabinet refers to the French king Louis XIV's military victories. A panel of marquetry showing the cockerel of France standing triumphant over both the eagle of the Holy Roman Empire and the lion of Spain and the Spanish Netherlands decorates the central door. On the drawer above the door, gilt-bronze military trophies flank a medallion portrait of Louis XIV. In the Dutch Wars of 1672 - 1678, France fought simultaneously against the Dutch, Spanish, and Imperial armies, defeating them all. This cabinet celebrates the Treaty of Nijmegen, which concluded the war. Two large figures from Greek mythology, Hercules and Hippolyta, Queen of the Amazons, representatives of strength and bravery in war, appear to support the cabinet.

The fleurs-de-lis on the top two drawers indicate that the cabinet was made for Louis XIV. As it does not appear in inventories of his possessions, it may have served as a royal gift. The Sun King's portrait appears twice on this work. The bronze medallion above the central door was cast from a medal struck in 1661 which shows the king at the age of twenty-one. Another medallion inside shows him a few years later.


Multiple relations

…. <ne ID=“ne127">a medallion portrait of Louis XIV</ne>.

……

<ne ID=“ne164”>this work</ne>.

<ne ID=“ne165”>The bronze medallion</ne> …..

<link href="coref.xml#id(ne165)" type="ident"> <anchor href="coref.xml#id(ne127)"/> </link>

<link href="coref.xml#id(ne165)" type=“poss"> <anchor href="coref.xml#id(ne164)"/> </link>


Link descriptors: type of relation, who-believes

GIVER: Do_you have <coref:de ID="de_20">diamond_mine.</coref:de>FOLLOWER: Yes I've got <coref:de ID="de_21">a gold_mine.</coref:de>GIVER: Ah. S--.FOLLOWER: ....GIVER: You don't have <coref:de ID="de_22">diamond_mine</coref:de> though.

FOLLOWER: No. <coref:de ID="de_23"> It </coref:de> is a gold_mine according to this one. Presumably <coref:de ID="de_24">that</coref:de> ‘s the same.GIVER: Well I've got <coref:de ID="de_25">a gold_mine</coref:de> as well you see.

<coref:link href="coref.xml#id(de_23)" who-believes="F" type="ident"> <coref:anchor href="coref.xml#id(de_22)"/></coref:link>


Markable descriptors: GNOME

<ne id="ne2" cat="poss-np" per="per3“ num="sing" gen="neut" gf="subj“ lftype="term“ onto="concrete" ani="inanimate" deix="deix-no" count="undersp-count" structure="undersp-structure" generic="generic-no" loeb="sem-function"> <ne id="ne3" cat="this-np" per="per3"num="sing“ gen="neut" gf="gen" lftype="term“ onto="concrete" ani="inanimate" deix="deix-yes" count="count-yes“ structure="atom" generic="generic-no" loeb=“disc-function"> This table's </ne> unusual materials and coloring …. </ne>


Descriptors, I: morpho-syntax

• Morpho-syntactic descriptors– (Syntactic) agreement features, syntactic

category, gramm function: • Indeed very useful both for interpretation

and for generation• But: separate level?• In GNOME: with markables for technical

reasons

– POS: separate level?


Descriptors, II: Semantic

• Semantic descriptors:– Wordsense: separate level?

– Abstractness, animacy, named entity: Reference?

– Countability: a difficult issue

– Collectiveneness, Inalienability (!)


Descriptors, III: Referential

• Referential descriptors:– Cardinality (!?)– Natural Gender (!)– Definiteness (?)– Informational status (!?)– Referential status (?)

• Not included, but very useful in GNOME:– LFTYPE (term / quantifier / predicate)


Link descriptors

• Referential target: “single target, use structural solutions for describing ambiguity” (see presentation)

• Objectal relations vs. linguistic relations– OK, but linguistic relations very difficult

(VENEX)


Ambiguity

15.12 M: we’re gonna take the engine E3

15.13 : and shove it over to Corning

15.14 : hook it up to the tanker car

15.15 : _and_

15.16 : send it back to Elmira

(from the University of Rochester TRAINS-91 dialogues)


Marking ambiguity

<de ID="de_01">we</de>'re gonna take <de ID="de_07"> the engine E3 </de> and shove <de ID="de_08"> it </de> over to <de ID="de_02">Corning</de>, hook <de ID="de_09"> it </de> up to <de ID="de_03">the tanker car</de>..._and_ and send <de ID="de_19">it</de> back to Elmira

<link href="coref.xml#id(de_08)" type="ident"> <anchor href="coref.xml#id(de_07)"/> </link>

<link href="coref.xml#id(de_19)" type="ident"> <anchor href="coref.xml#id(de_09)"/> <anchor href="coref.xml#id(de_03)"/> </link>


General questions

• What aspects of the proposal will become part of the standard? – What is the status of the recommendations re:

attributes? – For my own research I need richly annotated

corpora, but who else does? For which applications?

• How can the standard incorporate future extensions? (See next slides)– What about simplifications?


The difficult issues

• A `general purpose’ ?(coreference? anaphora?) annotation

• Ambiguity (GNOME, ARRAU)

• Underspecified relations (GNOME)

• Reference to abstract objects (ARRAU)

• Misunderstandings in multimodal reference (VENEX)


Object AND relation ambiguity

In 1740, Caffieri’s wife bought a royal privilege - a form of permit from the king – which allowed them to gild bronze as well as cast it; within the same workshop these two processes would usually have been done by separate businesses. After his son Philippe Caffieri joined the workshop in 1747,

they produced designs for chandeliers, ornaments for coaches, wall lights, and furniture mounts.


DE/REL AMBIGUITY

which allowed <ne id=“ne1”>them</ne> …. After <ne id=“ne2”>his son Philippe Caffieri</ne> ….joined <ne id=“ne3”>the workshop</ne> … <ne id=“ne4”>they</ne> produced …

<ante current=“ne4”> <anchor id=“ne1” rel=“ident”> <anchor id=“ne2” rel=“element”> <anchor id=“ne3” rel=“ident”></ante>


Product / instance metonymy & anaphora

XXX Cream

What you should know about XXX Cream

* What your cream does

XXX Cream is a strong and rapidly effective treatment for inflamed skin conditions such as eczema, psoriasis and dermatitis.

IT IS IMPORTANT TO READ THIS CAREFULLY BEFORE STARTING TREATMENT

Following this advice gives the cream a chance to work safely and effectively.

* What is in your cream

XXX Cream is one of a group of medicines called topical steroids. "Topical" means they are put on the skin.


Misunderstandings in the MapTask (VENEX)


MISUNDERSTANDINGS IN DIALOGUE

G037: Perfetto, allora questa... sei arrivata dove sta la punta delle barche? Perfect then, did you arrive at the top of the boats?F038: Sí YesG039: Mh, ora fai un mezzo cerchio. Now do a half-circleF040: La prima però, la prima figura <La seconda,> barche. The first however, the first landmark <the second> boatsG041: Quella più in alto. The one aboveF042: Ah, quella più <Eh> in alto? The one above?G043: Sí, ma non ci andare proprio sopra, cioé ci devi sta_, ci devi passà vicino non ci de_, non devi andá sulle figure. Fatto? Yes, but you should not go over it, I mean you must … you must get close but you cannot go over the landmark. Done?


MISUNDERSTANDINGS IN DIALOGUE, 2

F044: Sí YesG045: Perfetto, ora fai un mezzo cerchio verso destra come se volessi, come se volessi <Scendere.> fare ... Perfect, now draw a half-circle to the right as if you wanted to <go down>F046: G047: <Perché> No. No.F048: La figura barche sta più sopra di valle limpida. The boat landmark is above valle limpidaG049: Sta piú sopra? [It] is above?F050: Ce ne sono due, una piú sotto e una piú sopra. There is two, one below and one above


MISUNDERSTANDINGS IN DIALOGUE, CONT’D

G051: Mh, molto bene. Allora, mh, non c'hai barche un poco piú in alto spostato verso, verso destra, cioè non c'hai scritto valle limpida? Very good. Then, mh, don’t you have boats a bit higher towards, to the right, that is, don’t you have valle limpida F052: Sí, c'è scritto, peró poi piú sopra c'è un'altra figura barche. Yes, it’s there, however above that there is a second boats landmarkG053: Ci sono due figure barche? There are two boat landmarks?F054: Sí. Yes.


Conclusions

• Core aspects of the proposal good

• Need to clarify some of the more complex aspects and their role in the standard

• Would also be useful to expand on – Use in specific annotations / tasks– Relation with standoff– How future extensions can be included

1 Comments to: Referential Structures and Links (S. Salmon-Alt and L. Romary) Massimo Poesio University of Essex, Language & Computation Group.

Documents

isosigsem workshop

repr tilburg

links slide

descriptors of markables

markable descriptors

proposed descriptors

difficult venex slide

semantic semantic descriptors