Speech acts and Communicative Intentions for Urgency Detection

Proceedings of the 11th Joint Conference on Lexical and Computational Semantics, pages 289 - 298 July 14-15, 2022 ©2022 Association for Computational Linguistics Speech acts and Communicative Intentions for Urgency Detection Enzo Laurenti1 Nils Bourgon2 Farah Benamara2 1IJN, CNRS/ENS/EHESS, PSL University firstname.lastname@ens.fr Alda Mari1 Véronique Moriceau2 Camille Courgeon1 2 IRIT, Université de Toulouse, CNRS, Toulouse INP, UT3 firstname.lastname@irit.fr Abstract Recognizing speech acts (SA) is crucial for capturing meaning beyond what is said, mak- ing communicative intentions particularly rel- evant to identify urgent messages. This paper attempts to measure for the first time the im- pact of SA on urgency detection during crises, in tweets. We propose a new dataset annotated for both urgency and SA, and develop several deep learning architectures to inject SA into urgency detection while ensuring models gener- alisability. Our results show that taking speech acts into account in tweet analysis improves information type detection in an out-of-type configuration where models are evaluated in unseen event types during training. These re- sults are encouraging and constitute a first step towards SA-aware disaster management in so- cial media. 1 Introduction Discovered by (Austin, 1962) and extensively pro- moted by (Searle, 1975), speech acts (henceforth SA) have been the object of extensive discussion in the philosophical and the linguistic literature (Sadock, 2004; Portner, 2018). According to the Austinian initial view, SA are to achieve action rather than conveying information. When uttering I now pronounce you man and wife, the priest ac- complishes the action of marrying rather than just stating a proposition. Beyond these prototypical cases, the literature has quickly broaden the under- standing of the notion of SA as a special type of linguistic object that encompasses questions, or- ders and assertions and transcends propositional content revealing communicative intentions on the part of the speaker (Bach and Harnish, 1979; Port- ner, 2018; Giannakidou and Mari, 2021). Because recognizing speakers’ intentions is cru- cial for capturing meaning beyond what is said (Noveck, 2018), SA have given rise to an extensive body of work in the computational linguistics litera- ture where various approaches have been proposed to detect them in both synchronous (e.g., meeting, phone) (Stolcke et al., 2000; Keizer et al., 2002) as well as asynchronous dialogues (e.g., emails, tweet threads) (Carvalho and Cohen, 2005; Joty and Mohiuddin, 2018; Bracewell et al., 2012). SA have shown to be an important step in many down stream NLP applications such as strategic action prediction (Cadilhac et al., 2013), dialogue summa- rization (Goo and Chen, 2018) and conversational systems (Higashinaka et al., 2014). In this paper, we attempt to measure for the first time the role of SA on urgency detection in tweets, focusing on natural disasters (hurricanes, storms, floods, etc.). SA are particularly relevant to identify urgent messages, i.e. those that raise situational awareness over a crisis (including human/material damages, security instructions, etc.), providing therefore ac- tionable information that will help to set priorities for the human teams and decide appropriate rescue actions. By tweeting, speakers seek to achieve im- pact …

Content truncated to 500 words