A Cascade Model for Proposition Extraction in Argumentation

Abstract

We present a model to tackle a fundamental but understudied problem in computational argumentation: proposition extraction. Propositions are the basic units of an argument and the primary building blocks of most argument mining systems. However, they are usually substituted by argumentative discourse units obtained via surface-level text segmentation, which may yield text segments that lack semantic information necessary for subsequent argument mining processes. In contrast, our cascade model aims to extract complete propositions by handling anaphora resolution, text segmentation, reported speech, questions, imperatives, missing subjects, and revision. We formulate each task as a computational problem and test various models using a corpus of the 2016 U.S. presidential debates. We show promising performance for some tasks and discuss main challenges in proposition extraction.

Argument Mining: A Survey

“…Mood, anaphora, ellipsis, deixis, reported speech, and more all introduce new challenges for ADU identification. Jo et al (2019) have used a combination of techniques, some statistical, some rule-based, and some hybrid, organized in a cascade structure, in order to attempt to recover the propositional structure underlying ADUs, in order to improve the performance of other argument mining tasks.…”

Extracting Implicitly Asserted Propositions in Argumentation

“…Our work uses the data pre-processed by Jo et al (2019). This dataset has resolved anaphors in ADUs and paired ADUs with asserted propositions in a readily-available format 2 .…”

“…US2016: Our main data is Jo et al (2019)'s dataset of the 2016 U.S. presidential debates and commentary. We filtered 565 pairs of an ADU and its asserted proposition that are annotated with the following question types:…”

“…According to argumentation theory, the elementary units in argumentation are asserted propositions (Eemeren and Grootendorst, 1984). However, the dominant approach to extracting elementary units from textoften called argumentative discourse unit segmentation (Ajjour et al, 2017;Persing and Ng, 2016;Jo et al, 2019)-is rather simplistic and may even seem inconsistent with the theory. This approach segments text into smaller pieces (e.g., clauses) and treats each segment as an elementary unit of arguments.…”

“…More recently, a computational framework has been proposed to extract asserted propositions from ADUs (Jo et al, 2019). This cascade model proposes how to detect reported speech, questions, and imperatives, reconstruct any missing subjects, and make final revisions for grammar correction.…”