The Possible, the Plausible, and the Desirable: Event-Based Modality Detection for Language Processing

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 953–965 August 1–6, 2021. ©2021 Association for Computational Linguistics 953 The Possible, the Plausible, and the Desirable: Event-Based Modality Detection for Language Processing Valentina Pyatkin∗ Bar Ilan University pyatkiv@biu.ac.il Shoval Sadde∗ Bar Ilan University shovatz@gmail.com Aynat Rubinstein Hebrew University of Jerusalem aynat.rubinstein@mail.huji.ac.il Paul Portner Georgetown University paul.portner@georgetown.edu Reut Tsarfaty Bar Ilan University reut.tsarfaty@biu.ac.il Abstract Modality is the linguistic ability to describe events with added information such as how de- sirable, plausible, or feasible they are. Modal- ity is important for many NLP downstream tasks such as the detection of hedging, uncer- tainty, speculation, and more. Previous studies that address modality detection in NLP often restrict modal expressions to a closed syntac- tic class, and the modal sense labels are vastly different across different studies, lacking an ac- cepted standard. Furthermore, these senses are often analyzed independently of the events that they modify. This work builds on the theoreti- cal foundations of the Georgetown Gradable Modal Expressions (GME) work by Rubin- stein et al. (2013) to propose an event-based modality detection task where modal expres- sions can be words of any syntactic class and sense labels are drawn from a comprehensive taxonomy which harmonizes the modal con- cepts contributed by the different studies. We present experiments on the GME corpus aim- ing to detect and classify fine-grained modal concepts and associate them with their modi- fied events. We show that detecting and clas- sifying modal expressions is not only feasi- ble, but also improves the detection of modal events in their own right. 1 Introduction Modality refers to the linguistic ability to describe alternative ways the world could be.1 Modal ex- pressions aim to identify wishes, rules, beliefs, or norms in texts (Kratzer, 1981; Portner, 2009), which is a crucial part of Natural Language Under- standing (NLU) (Morante and Sporleder, 2012). Concretely, events in natural language are often reported in a manner that emphasizes non-actual perspectives on them, rather than their actual propo- sitional content. Consider examples (1a)–(1b): ∗Equal contribution 1In formal semantics, these alternatives are referred to as possible worlds or situations (Kripke, 1959; Lewis, 1973; Barwise and Perry, 1981; Kratzer, 2010). (1) a. We presented a paper at ACL’19. b. We did not present a paper at ACL’20. The propositional content p =“present a paper at ACL’X” can be easily verified for sentences (1a)- (1b) by looking up the proceedings of the confer- ence to (dis)prove the existence of the relevant pub- lication. The same proposition p is still referred to in sentences (2a)–(2d), but now in each one, p is described from a different perspective: (2) a. We aim to present a paper at ACL’21. b. We want to present a paper at ACL’21. c. We ought to present a paper at ACL’21. d. We are likely to present a paper at ACL’21. These sentences cannot be verified or falsified sim- ply by examining whether …

Content truncated to 500 words