Transcript Attention-centred Information in Language - ppt - celta

Attention-centred Information
in Language
MIC Sorbonne 2012
New Standards for Language Studies
Hélène WŁODARCZYK
Université Paris-Sorbonne (CELTA)
Attention:
from psychology to language
“Attention defines the mental ability to select stimuli, responses,
memories, or thoughts that are behaviourally relevant, among
the many others that are behaviourally irrelevant”.
(Corbetta, 1998, p. 831).
“it is likely that the most accessible (e.g., visually most salient)
referent will be articulated before other referents taking part in
the event and that it will be assigned as the most prominent
grammatical constituent, for example, the Subject.”
(Myachikov et al. 2009). From Russell Tomlin’s investigations.
See prof. Franz Stachowiak’s presentation at this workshop.
From Mental Representation to
Linguistic Utterance
Example of a mental representation (what a speaker means):
Semantic Situation : a man, a dog
Spatial Anchor: a park,
Temporal Anchor: the time of the speech act (present)
To build a linguistic utterance on the basis of a mental representation the speaker
may choose — in the limits of his/her own langage — a point of view on the
situation and propose or impose it on the hearer.
(1) The speaker chooses a verb to refer to the situation from a certain point of view
but this verb imposes its valence.
(2) The speaker then (a) chooses among the participants or anchors of the
situation his/her main centre of attention (CA); (b) assigns to the whole utterance
— or only to one distinguished CA contrasting with the rest — a meta-informative
old or new status.
Cultural and Individual Subjectivity
in Natural Language Utterances
1. Cultural choices are imposed on the speaker by his/her language
conventions including lexical means and grammaticized concepts
e.g.
• the use of honorifics or personal morphemes with a verb,
• the canonical word-order of subject, object and verb,
• the use of an article in each noun phrase,
• the use of one of the Tense or Aspect value available in a given language
etc.
2. The speaker’s individual choices (depending on his discourse strategy)
among the alternative constructions available in his/her language:
e.g.
•the choice of an active, passive or impersonal construction
•the choice of a verb among several lexemes refering to the same situation from
different points of view: sell or buy, give or receive,
etc.
The communicative field
common to hearer and speaker
In order to communicate information about his mental
representations, the speaker has to create a communicative
field making it possible to build an utterance which can be
understood by the hearer
In this field the speaker proposes to the hearer distinguished
chunks of information (his/her Centres of Attention worded as
Attention-Driven Phrases) and attaches to them old or new
meta-informative status.
Cf. Tomorrow Prof. Yasunari Harada and Dr Dorota Zielinska will give deeper
insight into the communicative field
What is Information ?
In the age of unification of cognitive sciences, the term
information should be used similarly in linguistics and in
information (computer) science.
We consider Information as the semantic content of an
utterance.
Information is produced when relations are established between
entities. The components of semantic situations are:
(1) static or dynamic frames (states, events and processes),
(2) their roles (enacted by animate agents and/or inanimate figures)
(3) and anchors (indicators of spatio-temporal relations).
In linguistic messages information is always partial:
speakers express only what is obligatory (grammaticized in
their language) and what is relevant from their point of view i.e.
what they pay attention to.
What is Meta-Information ?
Meta-Information is information about another information.
In order to achieve the ordering of non-linear mental
representations as texts (sequences of linguistic utterances),
the speaker must select one or more Centre(s) of Attention
(CA) and predicate about it/them.
Although meta-information belongs to the pragmatic
dimension of language it is not a mere supplement to the
semantic and syntactic structures of utterances.
What linguists usually call “information structure” following
the Prague School’s tradition we name meta-information.
Meta-Informative Old or New status
and Truth Validity of an Utterance
The semantic validation of utterances as True or False is well known in
logic and linguistics. But linguistic messages are always also characterised
by another type of validation : they have to be introduced into discourse
either as Old or New information, also called Given and New by linguists.
This need to introduce information as old or new gives rise to what we call
the meta-informative status of utterances.
We propose to distinguish thoroughly between information as a piece of
knowledge about a situation (as a representation of the world) and
predication as communication about it. Truth validity concerns information
itself whereas O/N meta-informative status concerns predication.
Many problems in linguistic science arise from the confusion of these two
levels.
Attention-Driven Phrases
Expressing Centres of Attention
In order to communicate non-linear semantic information the speaker has to
select one of the participants of the semantic situation as the global primary
centre of attention. And he/she may select another participant as a the
local secondary centre of attention.
A segment of an expression — representing an element of a semantic
situation or a whole situation — is “centered” (treated by the speaker as a
Centre of Attention) if it has been distinguished among other elements or
situations by linguistic meta-informative markers (syntactic, morphological,
prosodic or any pragmatic marker).
Centres of Attention
in the MIC theory
In the MIC theory, we treat Subject, Object, Topic and
Focus as resulting from the choice of a “Centre of Attention
(CA)” and consider them not only as psychological
phenomena but also as segments of linguistic utterances:
Attention-Driven Phrases (ADP).
No judgment may be made without selecting at least one
Centre of Attention (CA).
Centering is a structuring operation within the utterance —
not only within a text as in the American Centering Theory.
Grosz B. J., Joshi A.K., Weinstein S. (1995) “Centering: a framework for Modeling the
Local Coherence of Discourse”, Computational Linguistics, Vol. 21, Nr 2, 1995, p. 203-226
Meta-informative
and Semantic Patterns
The choice of a verb implies a point of view on the relation between the
participants of the situation. In Indo-European languages, when one of the
participant is an animate being or a human being acting as an agent it is
likely to be chosen as the main centre of attention (anthropocentric point of
view). A non human or non animate entity will be more frequently treated as
a secondary centre of attention.
Verb valency determines the canonical word order in the unmarked voice, i.
e. the active voice in nominative (non-ergative) languages.
Linguistic devices make it possible to change the CA of a situation:
Lexical voice transformation (passivization), makes it possible for the
speaker to choose which participant will be the global CA: to buy/to sell, to
give/to receive, donner/recevoir, kriegen/bekommen.
Grammatical devices: word-order, active, passive, or impersonal voice.
Base and Extended Utterances
duality of meta-informative status
Each utterance contains at least one centre of attention (CA).
The CA may be either of the same or of a different (Old or
New) meta-informative status than the rest of the utterance.
In a base utterance there is no contrast between the status of
the global CA and that of the rest of the utterance: it is either all
New or all Old.
Centres of attention of extended utterances contrast with the
rest of the utterance.
The Topic bearing an Old meta-informative status is in
contrast with the New Comment, the Focus of New metainformative status is in contrast with the Old Background (O).
Homogenous and contrasting
meta-informative statuses
Base Utterance Schemas
Base Utterance (Examples)
(New) Subject : (New) Predicate
#1 A new satellite has been
launched today.
(Old) Subject : (Old) Predicate
#2 Satellites turn around the Earth.
Extended Utterance Schemas
Extended Utterance
(Examples)
(Old) Topic : (New) Comment
#3 As regards satellite, X03 it has
been destroyed by a meteorite.
(New) Focus : (Old) Background
#4 It is satellite X03 which was
destroyed today.
Meta-informative pivots of discourse
Pragmatic levels
Meta-Level 1 : base Utterance
Meta-Level 2 : Extended Utterance
Meta-Level 3 : Dialogue/Text
Centres of Attention
Global
Local
Subject
Object
Topic
Focus
General
Theme
Particular
Theme
N.B. Level ø corresponds to the situation itself, it is not expressed directly
in linguistic utterances. In case of a binary relation, it can be represented
by the logical formula p(x, y) in which x enacts the active role and y the
passive role.
Patterns of Situations in Language:
Verbs and Noun Phrases
Verbs subcategorize (traditionally „govern”) NPs which point at different
dimensions of language : semantics and pragmatics. The verb valence is an
heterogenous bag of relations: the subject and direct object are Attentiondriven Phrases expressing centres of attention. Indirect objects (expressed
by PPs prepositional phrases or NPs in oblique cases) point at semantic
roles and anchors thanks to their case marker or to the preposition with
which they occur
The Subject and Direct Object as Centres of Attention combine freely with
semantic roles:
#1 A car (subject) hit a pedestrian.
#2 A pedestrian (subject) was hit by a car.
But there is a default relation between the subject of an active verb and the
active role, the direct object and the passive role in the semantic situation
pattern
#3 The doctor (subject) treated the patient (direct object) with antibiotics (indirect
Subject and Object
are Attention-Driven Phrases
Expressing Centres of Attention
The subject of the utterance is the most important and global AttentionDriven Phrase (ADP) about which the speaker predicates.
A secondary centre of attention may be expressed as the object, the
second ADP
“the subject and DO (direct object) may be viewed as the grammaticalized primary
and secondary topic of the discourse at the time when the clause in which they
partake is being processed.” (Givon 1994, 198)
Givón, T., 1994, "The Pragmatics of Voice : Functional and Typological Aspects" Givon, T.
(ed.), Voice and Inversion, Typological Studies in Language 28, Amsterdam-Philadelphia,
John Benjamins, pp. 3-44
N.B. what Givon calls “topic” we call more generally the “centre of attention”.
Attention- Driven Phrases and other NPs
in a Base Utterance
The difference between so-called abstract cases (Nom and Acc) and
concrete cases (Dat, Inst, Loc etc. ) appeared as a serendipity result in our
interactive investigation of the Polish gender with the SEMANA software.
This is shown on the next slide by the results of the STAT 3 analysis of the
Polish adjective declension morphemes (seen as syntactic relators).
WLODARCZYK André & Hélène (2008) & SAUVET Georges, "Morphological Data
Exploration - Using The Semana Platform" (Feature Granularity Problem in the Definition of
Polish Gender)", CASK Sorbonne 2008 (Language Data Mining) International conference,
June, 13th-14th, 2008, Université Paris-Sorbonne – Paris 4
This entire ppt is downloadable from the webpage http://celta.parissorbonne.fr/anasem/papers/
Axis 2 separates syntactic relators
PROJECTION DANS LE PLAN FACTORIEL [1,2]
|
Horizontal: Axe #2 (Inertie: 12.81%) ——— Vertical: Axe #1 (Inertie: 13.05%)
|
Largeur: 1.798197; Hauteur: 2.123853; Nombre de points : 27
+--------------------------------------------------+--------------------tem------------+--10
6
|
|
| 00
|
| axis 1
| 00
On one side:
|
|
| 00
|
|
| 00
 ta, to, ten, te*, ci, te| are only nomin. and/or accus.
|
| 00
tego may be either accusative
or
|
|
| 00
On the other side:
|
|
| 00
|
|
| 00
 tej, tych, temu, tymi,| ta*, tymi
|
tej
| 00
9
|
|
| 00
are only genitive, locative,
dative and/or instrum.
|
|
| 00
|
|
dat
| 00
|
|
| 00
|
sin|
| 00
|
te*
tego |
| 10
8 |
ta to ten 3
| 00
4
| 3 6
|
| 00
|
|
| 00
|
|
| 00
|
|
| 00
|
|
tym
ta*
| 00
3 | 00
|
|
2
1
|
|
| 00
|
|
| 00
|
|
| 00
+-----------------------------------------------inahum---gen---------------------------+--40
|
nhumas
| 20
| nom
acc
fem|
| 10
|
neu|
loc
| 00
|
|
| 00
|
|
| 00
|
|
| 00
|
|
| 00
|
|
ins
| 00
|
|
| 00
|
|
| 00
|
|
| 00
|
plu
| 00
|
|
| 00
|
ci
|
tych
| 10
|
te
|
| 00
19
16
|
|
| 00
|
|
| 00
|
|
| 00
|
|
| 00
|
|
tymi| 00
+--------------------------------------------------+-----------------------------------+--00
9
genitive
axis 2
How many Attention-Driven Phrases
are there in one Utterance ?
is it possible to express more than two centres of attention by more than two
attention-driven phrases in one utterance ?
There can be only two attention-driven phrases on the meta-informative level:
one subject and one (direct) object in one base utterance.
But the subject (or the object) can be a group of coordinated NPS:
(1) Peter and Mary bought fruits and candies.
Are there more than one local CA in
one utterance ?
In those sentences where two objects (direct and indirect) co-occur, or
where adverbial and/or prepositional NPs cooccur with direct objects, are
they cases of double Local CA? Are they to be regarded as members of a
single Local CA?
When there seems to be two local ADPs in addition to the Subject, there is
no doubt that one of them is either a topicalised or focalised phrase (the
CA of the meta-meta-informative level), i.e. neither a "second" nor “double”
subject, nor a "second" nor “double” object.
André Wlodarczyk (2007 in Japanese) revisited the question of “double”
subject in Japanese utterances and came to the conclusion that when
taking into acount the meta-meta-informative level a base utterance
entails only one subject.
The problem of double objects in some languages (German) remains to
be investigated in the MIC framework.
Are Indirect Objects
Attention-Driven Phrases
in Base Utterances ?
The question arises: when a verb governs more than one objects are they all
attention-driven NPs ?
Only the subject and direct object of a verb are attention-driven phrases, the
other NPs are indirect objects and adjuncts introduced by case markers or
prepositions pointing at the semantic role (or the type of anchor) of the entity
they refer to.
The question of give verbs in English
Bill gave a flower to his wife. -> A flower was given by Bill to his wife
Bill gave his wife a flower. -> She was given a flower by her husband.
The object which can be transformed into the subject of the verb in the
passive voice can be considered a direct object (the local CA).
Examples
(1) John opened the door with his key. --- is a base utterance: "the door" is
the Object, "with his key" is not meta-informatively centred.
(2a) John gave Mary a book. -- is a base utterance: "Mary" is an Object.
The phrase "book" is not meta-informatively centred (it is not an object
because it is not in the prime post verbal position).
(2b) John gave a book to Mary. -- is an extended utterance: "a book" is an
Object and "to Mary" is a Focus of the beneficiary role.
Combination of Centres of Attention
of the Meta- and Meta-Meta-Informative Level
Base utterance with two CAs: the subject and the object
(1)René Descartes published the Discourse on the Method in 1637 in
Leiden (Netherlands).
Extended utterance with three or four CAs : the subject and the object of
the meta-informative level ad the topic and/or focus of the meta-metainformative level can combine in one utterance
(2) As concerns the Discourse on the Method, the author, the French
philosopher René Descartes had to publish it abroad.
(3) It is in Leiden (Netherlands) that René Descartes first published the
Discourse on the Method.
(4) As concerns the Discourse on the Method, it is in Leiden (Netherlands)
that René Descartes first published it.
More than one Topic
but only one Focus ?
It is a well known fact that in spoken French it is possible to have several
topics one after the other:
(1)Moi, ma femme, sa voiture, elle est tombée en panne sur l’autoroute.
Lit. me, my wife,
her car, it broke down
on the motorway.
We found three utterances with two topics in a database of 580 Polish
utterances. However, we did not find in the same db examples with more
than one focus. This has to be verified on larger corpuses and in different
languages
CAs of the meta-meta-informative level
and their anaphorical resumption
One centre of attention may correspond to two attention-driven phrases in
the same utterance; this is the case in languages where attention-driven
phrases of the first meta-informative level (subject and object) are not
marked by a case morpheme and have to be referred to by an anaphoric
pronoun after they have been left-dislocated by topicalisation or
focalisation.
(1) The flowers, Peter bought them.
(2) It is Peter who bought the flowers.
In utterances (1) and (2) the CAs of the extended utterance are replaced
int the background or comment part by an anaphoric pronoun: who and
they.
In a language with case morphemes and explicit personal morpheme in
the verbform no anaphoric resumption is necessary e.g. in Polish:
(1 bis) Kwiaty kupił Piotr.
(2 bis) To Piotr kupił kwiaty.
Conclusion
We assume that there are only two attention-driven phrases on
the meta-informative level: one subject and one object in a
base utterance.
But, on the meta-meta-informative level, i.e. in an extended
utterance, it is possible to add new centres of attention : one, or
even more topics and one focus. It is doubtful whether there
can be more than one focus in an extended utterance.
For the time being it is necessary to verify the above
assumption on large corpuses and with native speaker
informants.
© Hélène WLODARCZYK