Challenges and Strategies in Cross-Cultural NLP

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 6997 - 7013 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Challenges and Strategies in Cross-Cultural NLP Daniel Hershcovich1 Stella Frank2 Heather Lent1 Miryam de Lhoneux1,3,4 Mostafa Abdou1 Stephanie Brandl1 Emanuele Bugliarello1 Laura Cabello Piqueras1 Ilias Chalkidis1 Ruixiang Cui1 Constanza Fierro1 Katerina Margatina5 Phillip Rust1 Anders Søgaard1 1University of Copenhagen 2University of Trento 3Uppsala University 4KU Leuven 5University of Sheffield dh@di.ku.dk Abstract Various efforts in the Natural Language Pro- cessing (NLP) community have been made to accommodate linguistic diversity and serve speakers of many different languages. How- ever, it is important to acknowledge that speak- ers and the content they produce and require, vary not just by language, but also by cul- ture. Although language and culture are tightly linked, there are important differences. Anal- ogous to cross-lingual and multilingual NLP, cross-cultural and multicultural NLP consid- ers these differences in order to better serve users of NLP systems. We propose a principled framework to frame these efforts, and survey existing and potential strategies. 1 A Framework for Cultural Awareness Language technology is rapidly advancing for a minority of the world’s languages. At the same time, the majority of languages are falling behind (Joshi et al., 2020). It is essential that language technology can serve the speakers of a wide variety of languages, who come from a wide variety of cultures. In this paper, we argue that doing so requires accommodating these speakers not only on a linguistic level, but also on a cultural level. Culture, like language, is a term that is hard to pin down, but generally describes the way of life of a collective group of people, and distinguishes them from other groups with other cultures (Mora, 2013; Shweder et al., 2007). Culture encompasses both material as well as non-material aspects, such as beliefs and linguistic practices (Kendall et al., 2005). Moreover, since “[c]ulture is the acquired knowledge people use to interpret experience and generate behavior” (Spradley, 1972), it is also the lens through which people understand linguistic messages. Culturally maladapted messages can 💬 Linguistic Form and Style ⛰ Common Ground 💙 Aboutness 🧭 Objectives and Values Figure 1: The role of culture in NLP, illustrated by four dimensions along which cultures vary, and for which NLP can be culturally biased: linguistic form and style, common ground, aboutness, and objectives (values). and will be misinterpreted; NLP must be culturally sensitive in order to avoid doing harm. Language and culture interact in a number of ways (see Figure 1). This paper aims to illuminate these connections, in order to motivate and inform future NLP work. Beyond linguistic form and style, how things are expressed in language, they include common ground, the shared knowledge based on which people reason and communicate; aboutness, what information people care to con- vey; and objectives or values, the goals people strive for (e.g., when developing language technol- ogy). Since language and culture are intertwined, the different dimensions …

Content truncated to 500 words