LESA: Linguistic Encapsulation and Semantic Amalgamation Based Generalised Claim Detection from Online Content

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 3178–3188 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 3178 LESA: Linguistic Encapsulation and Semantic Amalgamation Based Generalised Claim Detection from Online Content Shreya Gupta†∗, Parantak Singh‡∗, Megha Sundriyal†, Md Shad Akhtar†, Tanmoy Chakraborty† † IIIT-Delhi, India. ‡ Birla Institute of Technology and Science, Pilani, Goa, India. {shreyag, meghas, shad.akhtar, tanmoy}@iiitd.ac.in, f20170109@goa.bits-pilani.ac.in Abstract The conceptualization of a claim lies at the core of argument mining. The segregation of claims is complex, owing to the diver- gence in textual syntax and context across different distributions. Another pressing is- sue is the unavailability of labeled unstruc- tured text for experimentation. In this paper, we propose LESA, a framework which aims at advancing headfirst into expunging the for- mer issue by assembling a source-independent generalized model that captures syntactic fea- tures through part-of-speech and dependency embeddings, as well as contextual features through a fine-tuned language model. We re- solve the latter issue by annotating a Twit- ter dataset which aims at providing a testing ground on a large unstructured dataset. Ex- perimental results show that LESA improves upon the state-of-the-art performance across six benchmark claim datasets by an average of 3 claim-F1 points for in-domain experiments and by 2 claim-F1 points for general-domain experiments. On our dataset too, LESA outper- forms existing baselines by 1 claim-F1 point on the in-domain experiments and 2 claim- F1 points on the general-domain experiments. We also release comprehensive data annota- tion guidelines compiled during the annotation phase (which was missing in the current litera- ture). 1 Introduction The concept of a claim lies at the core of the ar- gument mining task. Toulmin (2003), in his argu- mentation theory, described the term ‘claim’ as ‘an assertion that deserves our attention’; albeit not very precise, it still serves as an initial insight. In recent years, Govier (2013) described a ‘claim’ as ∗ First two authors have equal contributions. The work was done when Parantak was an intern at LCS2 Lab, IIIT- Delhi. Text Claim? Alcohol cures corona. Yes Wearing mask can prevent corona. Yes Lord, please protect my family & the Philippines from the corona virus. No If this corona scare doesn’t end soon imma have to intervene No Table 1: A few examples of claim and non-claim. ‘a disputed statement that we try to support with reasons.’ The predicament behind the claim detection task exists given the disparity in conceptualization and lack of a proper definition of a claim. The task of claim detection across different domains has garnered tremendous attention so far owing to an uprise in social media consumption and by exten- sion the existence of fake news, online debates, widely-read blogs, etc. As an elementary exam- ple, claim detection can be used as a precursor to fact-checking; wherein segregation of claims aids in restricting the corpus that needs a fact-check. A few examples are shown in Table 1. Most of the existing works are built upon …

Content truncated to 500 words