tlt2005

tlt2005 - Experiments on Sense Annotations and Sense...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Experiments on Sense Annotations and Sense Disambiguation of Discourse Connectives Eleni Miltsakaki , Nikhil Dinesh , Rashmi Prasad , Aravind Joshi , and Bonnie Webber University of Pennsylvania Philadelphia, PA 19104 USA {elenimi,nikhild,rjprasad,joshi}@linc.cis.upenn.edu University of Edinburgh Edinburgh, EH8 9LW Scotland bonnie@inf.ed.ac.uk 1 Introduction Discourse connectives can be analyzed as discourse level predicates which project predicate-argument structure on a par with verbs at the sentence level. The Penn Discourse Treebank (PDTB) reflects this view in its design providing annotation of the discourse connectives and their arguments. Like verbs, discourse connectives have multiple senses. We present a set of manual sense annotation studies for three connectives whose arguments have been annotated in the PDTB. Using syntactic features computed from the Penn Treebank and a simple MaxEnt model, we have achieved some success in automatically disambiguating among their senses. 2 Background The Penn Discourse Treebank (PDTB) project [11] builds on basic ideas presented originally in Webber and Joshi 1998 [13] that connectives are discourse-level predicates which project predicate-argument structure on a par with verbs at the sentence level. In this framework, connectives are grouped into natural classes depending on how they project predicate-argument structure at the discourse level. The PDTB corpus includes annotations of four types of connectives: subordi- nating conjunctions, coordinating conjunctions, adverbial connectives and implicit connectives. 1 1 Official release of the annotated corpus is expected by November 2005. The final number of annotations in the corpus will amount to approximately 25K: 15K annotations covering 96 explicit 1 Because discourse connectives (like verbs) can be polysemous, the final version of the corpus will also have annotated the semantic role of each argument of each type of connective. This paper presents our work to date on manual and automated sense annotation of discourse connectives as predicates. 3 Sense annotations of connectives Senses can be distinguished or aggregated to a greater or lesser extent, depending on the needs of the application and the ability of annotators to distinguish them re- liably. As a result of initial annotation experiments, we have grouped senses of the connectives since , while and when into the following classes (1) temporal senses that are not causally (contingently) related, (2) contrastive senses, (3) contingent senses, and (4) senses that are simultaneously temporal and causal. Regarding temporal senses, we have not yet made finer distinctions [1]. The contrastive senses comprise comparative , oppositive and concessive senses, while the contingent senses comprise causal and conditional senses....
View Full Document

This note was uploaded on 03/06/2012 for the course CIS 630 taught by Professor Cis630 during the Spring '08 term at UPenn.

Page1 / 12

tlt2005 - Experiments on Sense Annotations and Sense...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online