Detecting Negation Cues and Scopes in Spanish

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 6902–6911 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 6902 Detecting Negation Cues and Scopes in Spanish Salud Mar´ıa Jim´enez-Zafra1, Roser Morante2, Eduardo Blanco3, Mar´ıa Teresa Mart´ın-Valdivia1, L. Alfonso Ure˜na-L´opez1 1SINAI, Computer Science Department, CEATIC, Universidad de Ja´en {sjzafra, maite, laurena}@ujaen.es 2CLTL Lab, Computational Linguistics, VU University Amsterdam r.morantevallejo@vu.nl 3University of North Texas eduardo.blanco@unt.edu Abstract In this work we address the processing of negation in Spanish. We first present a machine learning system that processes negation in Spanish. Specifically, we focus on two tasks: i) negation cue detection and ii) scope identification. The corpus used in the experimental framework is the SFU ReviewSP-NEG. The results for cue detection outperform state-of-the-art results, whereas for scope detection this is the first system that performs the task for Spanish. Moreover, we provide a qualitative error analysis aimed at understanding the limitations of the system and showing which negation cues and scopes are straightforward to predict automatically, and which ones are challenging. Keywords: negation processing, negation cues detection, scope identification, Spanish, SFU ReviewSP-NEG corpus 1. Introduction Negation is a phenomenon that “relates an expression e to another expression with a meaning that is in some way op- posed to the meaning of e” (Horn and Wansing, 2017). It is present in all languages and is always marked. Nega- tion has been studied from a theoretical perspective in vir- tually all languages, and determining the scope, that is, the part of the sentence affected by the negation cue, is critical to understanding sentences with negation (Ladusaw, 1992). Within computational linguistics, however, most previous work on negation targets English texts, albeit in a variety of domains. While negation processing systems are available for a few other languages, for Spanish the effort has focused mostly and only recently on annotation tasks. Yet Spanish is the second most widely spoken language in terms of na- tive speakers (Lewis and Gary, 2015) and the third most common language used on the Internet.1 Identifying negation cues and scopes in general—and in Spanish in particular—is challenging. First, negation cues may be discontinuous and ambiguous, i.e., the same se- quence of characters does not always indicate a negation. Second, scopes may span tokens before and after the nega- tion cue, and scopes do not need to cover full sentences or phrases. In the SFU ReviewSP-NEG corpus (Jim´enez- Zafra et al., 2018) most scopes (75.47%) cover less than 30% of the sentence they belong to. Third, two or more negations may be present in a sentence and their scopes may overlap. Ex. (1) and (2) illustrate some of these challenges with sentences selected from user-generated re- views of the SFU ReviewSP-NEG corpus.2 The first no in 1”Number of Internet Users by Language”, Internet World Stats, Miniwatts Marketing Group, 31 Decem- ber 2017, accessed 20 February 2019. URL: https: //www.internetworldstats.com/stats7.htm 2In the examples we mark in bold negation cues and enclose negation scopes between square brackets. …

Content truncated to 500 words