Evidently epistential adverbs are argumentative indicators: A corpus-based study



Musi, Elena ORCID: 0000-0003-2431-455X and Rocci, Andrea
(2017) Evidently epistential adverbs are argumentative indicators: A corpus-based study. Argument & Computation, 8 (2). 175 - 192.

[img] Text
aac_2017_8-2_aac-8-2-aac023_aac-8-aac023.pdf - OA Published Version

Download (340kB) | Preview

Abstract

Argumentative indicators of discourse relations constitute crucial cues for the mining of arguments. However, a comprehensive lexicon of these linguistic devices is so far lacking due to the scarcity of corpora argumentatively annotated and the absence of an empirically validated analytic methodology. Recent studies have shown that modals, that express that things might be otherwise, and evidentials, that point to the presence of information sources, are good candidates to work as argumentative indicators. On these grounds, we propose a systematic, non-language specific corpus-based procedure to identify indicators of argumentative discourse relations. We test the design of a multi-level annotation through the analysis of the English and Italian epistential adverbs evidently and evidentemente in comparable corpora of newspaper articles. We show that the annotation guidelines achieve consistent analytical results with expert annotators. Data analysis reveals that the two adverbs work as argumentative indicator both at the structural and at the inferential level: besides pointing to the presence of premises-conclusion relations, they recurrently pattern with causal argument schemes from the effect to the cause. The Italian adverb evidentemente is less polysemous and more frequent, thus working as a more reliable indicator.

Item Type: Article
Uncontrolled Keywords: Argumentative indicators, epistentials, multilevel annotation, newspaper articles
Depositing User: Symplectic Admin
Date Deposited: 08 Jul 2020 09:43
Last Modified: 08 Sep 2021 07:10
DOI: 10.3233/aac-170023
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3093172