Creating Local Coherence: An Empirical Assessment

20812

Authors

nan

Abstract

Two of the mechanisms for creating natural transitions between adjacent sentences in a text, resulting in local coherence, involve dis- course relations and switches of focus of at- tention between discourse entities. These two aspects of local coherence have been tradi- tionally discussed and studied separately. But some empirical studies have given strong evi- dence for the necessity of understanding how the two types of coherence-creating devices interact. Here we present a joint corpus study of discourse relations and entity coherence ex- hibited in news texts from the Wall Street Jour- nal and test several hypotheses expressed in earlier work about their interaction. 1

Keywords

nan

Full Content

Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the ACL, pages 313–316, Los Angeles, California, June 2010. c⃝2010 Association for Computational Linguistics Creating Local Coherence: An Empirical Assessment Annie Louis University of Pennsylvania Philadelphia, PA 19104, USA lannie@seas.upenn.edu Ani Nenkova University of Pennsylvania Philadelphia, PA 19104, USA nenkova@seas.upenn.edu Abstract Two of the mechanisms for creating natural transitions between adjacent sentences in a text, resulting in local coherence, involve dis- course relations and switches of focus of at- tention between discourse entities. These two aspects of local coherence have been tradi- tionally discussed and studied separately. But some empirical studies have given strong evi- dence for the necessity of understanding how the two types of coherence-creating devices interact. Here we present a joint corpus study of discourse relations and entity coherence ex- hibited in news texts from the Wall Street Jour- nal and test several hypotheses expressed in earlier work about their interaction. 1 Introduction Coherent discourse is characterized by local prop- erties that are crucial for comprehension. In fact, a long line of linguistics and computational linguistics tradition has proposed that several levels of struc- ture contribute to the creation of coherent discourse. Among these, the attentional structure (Grosz et al., 1995) and the relational structure (Mann and Thompson, 1988) of text, are the most widely dis- cussed in the literature. Centering theory (Grosz et al., 1995) is the dom- inant approach for describing and analyzing atten- tional structure. It assumes that readers of the text focus (center) their attention on a small number of salient discourse entities at a time and that there are preferred patterns for switching attention between entities mentioned in adjacent sentences. Relational This work was partially supported by NSF Grant IIS -07- 05671. structure theories, on the other hand, describe how certain discourse (also called rhetorical or coher- ence) relations such as CONTRAST and CAUSE are inferred by the reader between adjacent units of text. The existence of richly annotated corpora and the development of automatic applications based on the theories have allowed empirical assessments of the validity and generality of these theories individually. Such work has also provided increasingly strong evidence that attentional and relational structures need to be taken into account simultaneously. The motivation behind such proposals have been the em- pirical findings that “weak” discourse relations such as ELABORATION are the most common type of re- lations, and that a large percentage of adjacent sen- tences in fact do not have any entities in common. In particular, a corpus based evaluation of Center- ing theory found that only 42% of pairs of adjacent sentences have at least one entity in common and hypothesized that discourse relations are responsible for creating local coherence in the remaining cases (Poesio et al., 2004). At the same time, the work of Knott et al. (2001) has discussed several theo- retical complications arising from the existence of the very common and semantically weak ELABO- RATION relation. These researchers propose that re- placing …

Content truncated to 500 words

Labeled Text

Two of the mechanisms for creating natural transitions between adjacent sentences in a text, resulting in local coherence, involve discourse relations and switches of focus of attention between discourse entities. These two aspects of local coherence have been traditionally discussed and studied separately. But some empirical studies have given strong evidence for the necessity of understanding how the two types of coherence-creating devices interact. Here we present a joint corpus study of discourse relations and entity coherence exhibited in news texts from the Wall Street Journal and test several hypotheses expressed in earlier work about their interaction. 1

Publication Details

Paper ID:
20812
Year:
2010
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan