Casting Implicit Role Linking as an Anaphora Resolution Task

26333

Authors

nan

Abstract

Linking implicit semantic roles is a challeng- ing problem in discourse processing. Unlike prior work inspired by SRL, we cast this prob- lem as an anaphora resolution task and embed it in an entity-based coreference resolution (CR) architecture. Our experiments clearly show that CR-oriented features yield strongest performance exceeding a strong baseline. We address the problem of data sparsity by apply- ing heuristic labeling techniques, guided by the anaphoric nature of the phenomenon. We achieve performance beyond state-of-the art. 1

Keywords

nan

Full Content

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 1–10, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics Casting Implicit Role Linking as an Anaphora Resolution Task Carina Silberer∗ School of Informatics University of Edinburgh Edinburgh, UK c.silberer@ed.ac.uk Anette Frank Department of Computational Linguistics Heidelberg University Heidelberg, Germany frank@cl.uni-heidelberg.de Abstract Linking implicit semantic roles is a challeng- ing problem in discourse processing. Unlike prior work inspired by SRL, we cast this prob- lem as an anaphora resolution task and embed it in an entity-based coreference resolution (CR) architecture. Our experiments clearly show that CR-oriented features yield strongest performance exceeding a strong baseline. We address the problem of data sparsity by apply- ing heuristic labeling techniques, guided by the anaphoric nature of the phenomenon. We achieve performance beyond state-of-the art. 1 Introduction A widespread phenomenon that is still poorly stud- ied in NLP is the meaning contribution of unfilled semantic roles of predicates in discourse interpreta- tion. Such roles, while linguistically unexpressed, can often be anaphorically bound to antecedent ref- erents in the discourse context. Capturing such im- plicit semantic roles and linking them to their an- tecedents is a challenging problem. But it bears im- mense potential for establishing discourse coherence and for getting closer to the aim of true NLU. Linking of implicit semantic roles in discourse has recently been introduced as a shared task in the SemEval 2010 competition Linking Events and Their Participants in Discourse (Ruppenhofer et al., 2009, 2010). The task consists in detecting un- filled semantic roles of events and determining an- tecedents in the discourse context that these roles ∗ The work reported in this paper is based on a Master’s Thesis conducted at Heidelberg University (Silberer, 2011). can be understood to refer to. In (1), e.g., the pred- icate jealousy introduces two implicit roles, one for the experiencer, the other for the object of jealousy involved. These roles can be bound to Watson and the speaker (I) in the non-local preceding context. (1) Watson won’t allow that I know anything of art but that is mere jealousy because our views upon the subject differ. (2) IReader was sitting reading in the chairP lace. In contrast to implicit roles that can be discourse- bound to an antecedent as in (1), roles can be inter- preted existentially, as in (2), with an unfilled TEXT role of the READING frame that cannot be anchored in prior discourse. The FrameNet paradigm (Fill- more et al., 2003) that was used for annotation in the SemEval task classifies these interpretation dif- ferences as definite (DNI) vs. indefinite (INI) null instantiations (NI) of roles, respectively. 2 Implicit Role Reference: A Short History Early studies. The phenomenon of implicit role re- ference is not new. It has been studied in a number of early approaches. Palmer et al. (1986) treated un- filled semantic roles as special cases of anaphora and coreference resolution (CR). Resolution was guided by domain knowledge encoded in a knowledge- based system. Similarly, Whittemore et …

Content truncated to 500 words

Labeled Text

Linking implicit semantic roles is a challenging problem in discourse processing. Unlike prior work inspired by SRL, we cast this problem as an anaphora resolution task and embed it in an entity-based coreference resolution (CR) architecture. Our experiments clearly show that CR-oriented features yield strongest performance exceeding a strong baseline. We address the problem of data sparsity by applying heuristic labeling techniques, guided by the anaphoric nature of the phenomenon. We achieve performance beyond state-of-the art. 1

Publication Details

Paper ID:
26333
Year:
2012
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan