Attribution and the (non-)alignment of syntactic and discourse arguments of connectives

Abstract

The annotations of the Penn Discourse Treebank (PDTB) include (1) discourse connectives and their arguments, and (2) attribution of each argument of each connective and of the relation it denotes. Because the PDTB covers the same text as the Penn TreeBank WSJ corpus, syntactic and discourse annotation can be compared. This has revealed significant differences between syntactic structure and discourse structure, in terms of the arguments of connectives, due in large part to attribution. We describe these differences, an algorithm for detecting them, and finally some experimental results. These results have implications for automating discourse annotation based on syntactic annotation.

A Constituent-Based Approach to Argument Labeling with Joint Inference in Discourse Parsing

“…Next, a tree subtraction algorithm was used to extract the arguments. However, as pointed out in Dinesh et al (2005), it is not necessarily the case that a connective, Arg1, or Arg2 is dominated by a single node in the parse tree (that is, it can be dominated by a set of nodes). Figure 1 .…”

“…In comparison, labeling full argument spans can provide a complete solution to argument labeling in discourse parsing and has thus attracted increasing attention recently, adopting either a subtree extraction approach (Dinesh et al (2005), Lin et al (2014)) or a linear tagging approach (Ghosh et al (2011)). …”

“…As a representative subtree extraction approach, Dinesh et al (2005) proposed an automatic tree subtraction algorithm to locate argument spans for intra-sentential subordinating connectives. However, only dealing with intra-sentential subordinating connectives is not sufficient since they constitute only 40.93% of all cases.…”

Hybrid Approach to PDTB-styled Discourse Parsing for CoNLL-2015

“…We adopted Dinesh et al (2005)'s tree subtraction method for subordinating conjunctions. This method takes a constituent parse tree as an input and detects argument spans as follows:…”

Because We Say So

“…In terms of Segmented Discourse Representation Theory (SDRT) (Asher and Lascarides, 2003), for example, the relation QUESTION-ANSWER-PAIR in (1) should be taken to hold only between α and β; the content of the matrix clause should be likewise excluded from the second argument of EX-PLANATION in (2) (Hunter et al, 2006). Similarly, the Penn Discourse Treebank (PDTB) would relate only α and β in (2) with implicit because (Dinesh et al, 2005;Prasad et al, 2006).…”