Speaker Attitudes Detection through Discourse Markers Analysis
Journal
Deep Learning and Neural Approaches for Linguistic Data
Date Issued
2021-09-30
Author(s)
Valūnaitė Oleškevičienė, Giedrė
Liebeskind, Chaya
Silvano, Purificação
Chiarcos, Christian
Damova, Mariana
Abstract
Speaker attitude detection is important for processing opinionated text. Survey data as such provide a valuable source of information and research for different scientific disciplines. They are also of interest to practitioners such as policymakers, politicians, government bodies, educators, journalists, and all other stakeholders with occupations related to people and society. Survey data provide evidence about particular language phenomena and public attitudes to provide a broader picture about the clusters of social attitudes. In this regard, attitudinal discourse markers play a central role in the sense that they are pointers to the speaker's attitudes. These single word or multiword expressions (MWE) are mainly drawn from syntactic classes of conjunctions, adverbials, and prepositional phrases (Fraser, 2009), as well as expressions such as you know, you see, and I mean (Schiffrin, 2001; Hasselgren, 2002; Maschler & Schiffrin, 2015). Discourse markers are regarded as significant discourse relations’ triggers, and, consequently, are largely studied (eg Sanders et al. 1992; Knott & Dale 1994; Wellner et al 2006; Taboada & Das 2013; Das 2014; Das & Taboada 2019; Silvano 2011). Recently, discourse relations and discourse marker research has gained certain impetus with corpora annotation for exploring discourse structure in texts, for example, RST-DT English corpus (Carlson, Marcu & Okurowski 2003); Penn Discourse Tree Bank (PDTB)(Prasad et al. 2008); SDRT Annodis French corpus (Afantenos et al., 2012).
Subjects
File(s)![Thumbnail Image]()
Loading...
Name
Workshop_on_Deep_Learning_and_Neural_App.pdf
Size
457.69 KB
Format
Adobe PDF
Checksum
(MD5):cb4612ddb8a549ccbb6f9672099664df