Probabilistic Head-Driven Parsing for Discourse Structure

Proceedings of the 9th Conference on Computational Natural Language Learning (CoNLL), pages 96–103, Ann Arbor, June 2005. c⃝2005 Association for Computational Linguistics Probabilistic Head-Driven Parsing for Discourse Structure Jason Baldridge and Alex Lascarides School of Informatics University of Edinburgh 2 Buccleuch Place Edinburgh, EH8 9LW Scotland, UK {jbaldrid,alex}@inf.ed.ac.uk Abstract We describe a data-driven approach to building interpretable discourse structures for appointment scheduling dialogues. We represent discourse structures as headed trees and model them with probabilis- tic head-driven parsing techniques. We show that dialogue-based features regard- ing turn-taking and domain specific goals have a large positive impact on perfor- mance. Our best model achieves an f- score of 43.2% for labelled discourse rela- tions and 67.9% for unlabelled ones, sig- nificantly beating a right-branching base- line that uses the most frequent relations. 1 Introduction Achieving a model of discourse interpretation that is both robust and deep is a major challenge. Consider the dialogue in Figure 1 (the sentence numbers are from the Redwoods treebank (Oepen et al., 2002)). A robust and deep interpretation of it should resolve the anaphoric temporal description in utterance 154 to the twenty sixth of July in the afternoon. It should identify that time and before 3pm on the twenty- seventh as potential times to meet, while ruling out July thirtieth to August third. It should gracefully handle incomplete or ungrammatical utterances like 152 and recognise that utterances 151 and 152 have no overall effect on the time and place to meet. According to Hobbs et al. (1993) and Asher and Lascarides (2003), a discourse structure consisting of hierarchical rhetorical connections between utter- ances is vital for providing a unified model of a wide 149 PAM: maybe we can get together, and, discuss, the planning, say, two hours, in the next, couple weeks, 150 PAM: let me know what your schedule is like. 151 CAE: okay, let me see. 152 CAE: twenty, 153 CAE: actually, July twenty sixth and twenty seventh looks good, 154 CAE: the twenty sixth afternoon, 155 CAE: or the twenty seventh, before three p.m., geez. 156 CAE: I am out of town the thirtieth through the, 157 CAE: the third, I am in San Francisco. Figure 1: A dialogue extract from Redwoods. range of anaphoric and intentional discourse phe- nomena, contributing to the interpretations of pro- nouns, temporal expressions, presuppositions and ellipses (among others), as well as influencing com- municative goals. This suggests that a robust model of discourse structure could complement current ro- bust interpretation systems, which tend to focus on only one aspect of the semantically ambiguous ma- terial, such as pronouns (e.g., Str¨ube and M¨uller (2003)), definite descriptions (e.g., Vieira and Poe- sio (2000)), or temporal expressions (e.g., Wiebe et al. (1998)). This specialization makes it hard to assess how they would perform in the context of a more comprehensive set of interpretation tasks. To date, most methods for constructing discourse structures are not robust. They typically rely on grammatical input and use symbolic methods which inevitably lack coverage. One …

Content truncated to 500 words