Interpretation and Discourse Structure

advertisement
Interpretation and Discourse Structure
Klaus von Heusinger & Edgar Onea
1. Synopsis

Motivation
◦ Project related observations:
▪ Results of project B1
 the semantic structure of -ung nominalizations has been analyzed at the
level of syntax-semantics interface,
Observation:
◦ discourse context is strongly involved in the process of disambiguation:
▪ causal relations
▪ temporal relations
▪ Results of project C2
 semantic and pragmatic conditions for differential object marking have been
analysed.
Observation:
◦ activation status (referential category) is relevant for differential object
marking
◦ topicality is relevant for case marking
◦ salience related operations are relevant for case marking
◦ General observations:
▪ Discourse, structured in terms of rhetorical relations plays a crucial role in the
interpretation of linguistic expressions:
 the resolution of anaphora
 the establishment of temporal and causal relations between event
 the projection properties of presuppositions
 the derivation of Gricean implicatures etc.
▪ Salience and topicality play a crucial role
 in anaphora resolution
 in the choice of referential expressions
 in discourse coherence

Outline of the project
◦ The aim of the project is to explore the enrichment of discourse representation
theories with principled notions like salience and topicality and analyze grammatical
effects related to this field:
◦ The phenomena to be discussed are:
▪ Anaphora resolution in the case of different types of anaphoric expressions
▪ The interaction of discourse structure and grammatical markers in anaphora
resolution
▪ The role of topicality and topic shifting in anaphora resolution and the
construction of SDRS.
◦ General claim:
▪ Discourse structure in terms of rhetorical relations and discourse structure in
terms of activation and salience interact in the process of anaphora resolution
▪ Anaphora resolution not only depends on the discourse context but also
influences the choice of the discourse update
▪ Topicality is related both to activation and salience on the one hand and
important in the construction of discourse representation and anaphora
resolution.
◦ Aims
▪ Principle aim is to gain a deeper understanding of conventional and contextual
factors involved in anaphora resolution
▪ Secondary aim is to gain a better understanding of the role topicality interacts
1
with discourse structure.
2. Background: Rhetorical relations and discourse structure

It has been argued that discourse representation involves rhetorical relations (cf. e.g.
Asher & Lascarides 2003).
Discourse is a complex representation structure involving different levels of
embedding and different discourse links (rhetorical relations):
◦
(1)
a. Peter kissed Mary today.
b. He was waiting for this moment for a long time.
c. But she never let him kiss her.
d. She changed her mind, only after her boyfriend, Jack, left her.
e. Afterwards, Jack saw them in the park.
f. They looked pretty happy.
narration
a
e
explanation
elaboration
elaboration
elaboration
b
c
contrast
◦
◦

d
f
contrast
The exact rhetorical relation can be conventionally marked (but, because, for
example).
▪ Or it can be underspecified.
▪ The discourse representation segment to which a new segment will attach is
often also underspecified (exception e.g. syntactic subordination).
 One task of discourse interpretation is to “guess” how to update the
discourse with a new utterance.
Rhetorical relations come with truth conditional effects.
▪ In narration t1 < t2
▪ In explanation t2 <= t1 etc.
Rhetorical relations play a role in a number of phenomena:
◦
Temporal order
▪ The inferences regarding temporal order depend on the rhetorical relations.
(2)
a. John fell. Jack pushed him.
b. John fell. Jack helped him get up.
◦
(3)
(explanation: t1 > t2)
(narration: t1 < t2)
Lexical disambiguation (enrichment)
▪ Discourse context can determine the interpretation of underspecified expressions
beyond what may be a conventional part of a generative lexicon.
a. John started a book.
b. ? John started with the kitchen.
c. John cleaned the house yesterday. He started with the kitchen. He went on to the
bathroom and finished with the living room.
2
◦
(4)
Implicatures
▪ Rhetorical relations can help deriving implicatures.
A: Smith doesn't seem to have a girlfriend
B: He's been paying lots of visits to New York lately.
▪
◦
(5)
Assuming that the answer of B is interpreted as counterevidence, given general
world knowledge it can be inferred that B means that Smith has a girlfriend in
New York.
 Crucially, this implicature has been derived without reasoning about belief
states of the speaker.
 The inference that B's utterance establishes the rhetorical relation of
counterevidence is signaled by prosody.
Presupposition projection
Either John didn't solve the problem or else Mary realized that the
problem has been solved.
▪
∂ The problem has been solved.
 Intuitively narrow scope correct: Either John didn't solve the problem or else
(the problem has been solved) and Mary realized it.
 In context SDRT can account for the narrow scope reading by maximizing
discourse coherence.
3. Anaphora resolution (background)

(6)
It is known that are constraints on anaphora resolution
◦ Accessibility (DRT)
▪ An anaphoric expression cannot pick a non-accessible antecedent.
?? If the shei is alone, a girli is happy.
◦
(7)
Centering theory
▪ There are strict hierarchies of salience which determine the most likely
resolution of anaphoric expressions:
John wanted Tony to join him on a sailing expedition.
He called him at 6 AM.
He was sick.
◦
Salience/accessibility
▪ The antecedent to an anaphoric pronoun must be sufficiently activated
▪ Otherwise anaphoric definite descriptions may re-activate it.
(8)
A small bird1 came toward the skiff from the north. He1 was a warbler and flying very
low over the water. The old man could see that he1 was very tired. The bird1 made the
stern of the boat and rested there. Then he1 flew around the old man’s head and rested
on the line where he1 was more comfortable.
◦
Discourse structure
▪ In (2f) it cannot refer to the salmon because of boundary constraints.
(9)
a. Max had a great evening last night.
b. He had a great meal.
c. He ate salmon.
d. He devoured lots of cheese.
3
e. He then won a dancing competition.
f. ?? It was a beautiful pink.
4. Additional aspects of anaphora resolution

SDRT helps predicting possible anaphora resolution, however it needs to be enhanced
with a theory of salience:
◦ SDRT would predict that in (10f) she can access his wife in (10a), but in such a case
his wife is not sufficiently activated.
(10)
a. Max had a great evening last night with his wife.
b. He had a great meal.
c. He ate salmon.
d. He devoured lots of cheese.
e. He then had a very friendly conversation with his father.
f. (?She)/His wife finally took him to a dancing competion.
◦
In (11f) she can refer back to the girl because of the higher level of activation.
(11)
a. Max had a great evening last night with a really gorgeous young girl from Mexico.
b. He had a great meal.
c. He ate salmon.
d. He devoured lots of cheese.
e. He then had a very friendly conversation with his father.
f. She/The girl finally took him to a dancing competion.

(12)
The notion of topic and topic shift also needs to be considered for anaphora resolution:
◦ In (12d) er can refer back to the waiter because it is in a topic shifting construction.
If “aber” is ommitted or placed elsewhere in (12d) the reading that er refers to the
waiter is much less probable.
a. Der Kellner eilte zum Gast.
b. Der hatte sich nämlich gemeldet.
c. Er wollte ein Bier.
d. Er aber sagte, dass Bier ungesund sei.
◦
Topicality can be signaled at the sentence-level, and it can help establish the right
discourse structure.
▪ Jasinskaja (2008) argues that given two utterances without any overt marking of
the rhetorical relation, elaboration or explanation are the most likely discourse
relations, as long as:
 The topic does not change.
 The discourse update is exhaustive.
(13)
a. Fred broke the carafe.
b. He hit it against the sink.
c. As for John, he hit it against the sink.
Elaboration (a,b)
Narration, Enumeration (a,c)
5. Enriching SDRT


We argue that discourse structured in terms of rhetorical relations is relevant for
anaphora resolution.
We argue, however that in addition, discourse structure must be enriched by a theory
of salience and topicality.
4
The general plan is to consider different levels of hierarchic discourse representation
and integrate salience hierarchies on them as well as to determine the role of
discourse relation in the activation and deactivation of discourse referents.
◦ To integrate a notion of topic and topic shift that helps not only in anaphora
resolution but also in determining the correct discourse relations (and which is also
compatible with more general notions of topicality, contrastive topics etc.)
The theoretical aim of the project goes, however, beyond this scope: we also wish to
investigate to which extent grammatical phenomena such as topic markers, the choice
of referential expressions etc. can conventionally interact with such an enriched
discourse representation structure.
◦

6. Topic shifters – a specific example

(14)
(15)
Topic shift seems to be generally relevant for discourse relations and is also
grammaticalized in languages like German or Hungarian
◦ In German, topic shift is argued to be triggered by the presence of connectives in
the immediate pre-verbal position but after a topic.
Peter hat versucht, einen Apfel zu stehlen.
John aber war wieder einmal aufmerksam.
Peter hat versucht, einen Apfel zu stehlen.
John nämlich war wieder einmal hungrig.
▪
One possible test for this phenomenon is co-reference:
(16)
a. Peteri sieht traurig aus. Der Jungei/j ist nämlich krank.
b. Peteri sieht traurig aus. Der Junge*i/j nämlich ist krank.
▪
(17)
(18)

Topic shift can have additional impact on discourse inferences. Consider the
contrast between (23) and (25). There is a pragmatic contrast, but at the time
beeing we are not able to capture it in a precise manner.
 It seems however, that (23) is more appropriately paraphrased in (26) than
(25) cannot.
Peter hat versucht, einen Apfel zu stehlen.
John war nämlich wieder einmal hungrig.
Peter hat versucht, einen Apfel zu stehlen, wegen John.
In addition we wish to investigate also possible interaction with more syntactic oriented
frameworks such as binding theory with regard to anaphora resolution.
◦ Specifically in the analysis of topic shifters in German (and other languages) we will
also look a syntactic properties of such phenomena.
Previous work for C3
Bücher and edit. Vol.

von Heusinger, Klaus 1999. Intonation and Information Structure. The Representation
of Focus in Phonology and Semantics. Habilitationsschrift. Universität Konstanz.

von Heusinger, Klaus & Schwabe, Kerstin (eds.) 2001. Proceedings of the Workshop
„Information Structure and the Referential Status of Linguistic Expressions“. ZAS Papers
in Linguistics (ZASPiL) 23. Zentrum für Allgemeine Sprachwissenschaft (ZAS), Berlin.

von Heusinger, Klaus & Schwabe, Kerstin (eds.) 2001. Special Issue „NP-Interpretation
and Information Structure“ Theoretical Linguistics 27.2/3.

von Heusinger, Klaus & Schwabe, Kerstin (eds.) 2002. Special Issue „Information
Structure and Sentence Type“ Theoretical Linguistics 28.1.
5
Papers

von Heusinger, Klaus 2000. Anaphora, Antecedents, and Accessibility. Theoretical
Linguistics 26, 75-93.

von Heusinger, Klaus 2001. Information Structure and Discourse Semantics. The Prague
Bulletin of Mathematical Linguistics 73-74, 13-32.

von Heusinger, Klaus 2002. Information Structure and the Partition of Sentence
Meaning. In: E. Hajicová & P. Sgall & J. Hana & T. Hoskovec (eds.). Travaux du Cercle
Linguistique de Prague n. s. / Prague Linguistic Circle Papers 4. Amsterdam;
Philadelphia: Benjamins, 275-305.

von Heusinger, Klaus 2003. The Double Dynamics of Definite Descriptions. In: J.
Peregrin (ed.). Meaning in the Dynamic Turn. Amsterdam: Elsevier, 150-168.

von Heusinger, Klaus 2004. Focus Particles, Sentence Meaning, and Discourse
Structure. In: W. Abraham & A. ter Meulen (eds.). Composing Meaning. Amsterdam:
Benjamins, 167-193

von Heusinger, Klaus 2006. Salience and anaphoric definite noun phrases. Acta
lingvistica hafniensia, Theme issue. Explorations in the semantics-pragmatics interface.
Guest editors: Maj-Britt Mosegaard Hansen (U. of Copenhagen) and Ken Turner (U. of
Brighton), 33-53.

von Heusinger, Klaus 2007. Alternative Semantics for Definite NPs. In: K. Schwabe & S.
Winkler (eds.). On Information Structure , Meaning and Form. Amsterdam: Benjamins,
485-508

von Heusinger, Klaus 2007. Discourse Structure and Intonational Phrasing. In: D.
Büring & M. Gordon & Ch. Lee (eds.). Topic and Focus: Intonation and Meaning.
Theoretical and Crosslinguistic Perspectives. Dordrecht: Springer, 265-290.

von Heusinger, Klaus 2007. Accessibility and definite noun phrases. In: M. Schwarz & M.
Consten & M. Knees (eds.). Anaphors in Text: Cognitive, Formal and Applied
Approaches to Anaphoric Reference. Dordrecht: Benjamins, 123-144.
6
Download