VivesDebate-Speech: A Corpus of Spoken Argumentation to Leverage Audio Features for Argument Mining

TL;DR

The paper tackles the challenge of analyzing spoken argumentation by incorporating audio cues into argument mining. It introduces VivesDebate-Speech, a large public corpus built from the VivesDebate dataset with audio, BIO annotations, and word-level timestamps, enabling joint audio-text ADU segmentation and classification research. Two modeling approaches are explored—an end-to-end BIO tagging model and a cascaded segmentation-classification pipeline—showing that audio features improve ADU segmentation and overall macro-F1 scores, with best results around 0.51 on the test set. This work provides a valuable multimodal resource and baseline methodology that can drive future research in speech-based argument mining and related NLP tasks.

Abstract

In this paper, we describe VivesDebate-Speech, a corpus of spoken argumentation created to leverage audio features for argument mining tasks. The creation of this corpus represents an important contribution to the intersection of speech processing and argument mining communities, and one of the most complete publicly available resources in this topic. Moreover, we have performed a set of first-of-their-kind experiments which show an improvement when integrating audio features into the argument mining pipeline. The provided results can be used as a baseline for future research.

Figures (2)

• Figure 1: Overview of the proposed cascaded approach.
• Figure 2: Dev set F1 score as a function of maximum segment length (s), SHAS-multi segmenter followed by text classifier.