Identifying Cross-Cultural Differences in Word Usage

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 674–683, Osaka, Japan, December 11-17 2016. Identifying Cross-Cultural Differences in Word Usage Aparna Garimella and Rada Mihalcea University of Michigan {gaparna,mihalcea}@umich.edu James Pennebaker University of Texas at Austin pennebaker@utexas.edu Abstract Personal writings have inspired researchers in the fields of linguistics and psychology to study the relationship between language and culture to better understand the psychology of people across different cultures. In this paper, we explore this relation by developing cross-cultural word mod- els to identify words with cultural bias – i.e., words that are used in significantly different ways by speakers from different cultures. Focusing specifically on two cultures: United States and Aus- tralia, we identify a set of words with significant usage differences, and further investigate these words through feature analysis and topic modeling, shedding light on the attributes of language that contribute to these differences. 1 Introduction According to Shweder et al. (1998), “to be a member of a group is to think and act in a certain way, in the light of particular goals, values, pictures of the world; and to think and act so is to belong to a group.” Culture can be defined as any characteristic of a group of people, which can affect and shape their beliefs and behaviors (e.g., nationality, region, state, gender, or religion). It reflects itself in people’s everyday thoughts, beliefs, ideas, and actions, and understanding what people say or write in their daily lives can help us understand and differentiate cultures. In this work, we use very large corpora of personal writings in the form of blogs from multiple cultures1 to understand cultural differences in word usage. We find inspiration in a line of research in psychology that poses that people from different cultural backgrounds and/or speaking different languages perceive the world around them differently, which is reflected in their perception of time and space (Kern, 2003; Boroditsky, 2001), body shapes (Furnham and Alibhai, 1983), or surrounding objects (Boroditsky et al., 2003). As an example, consider the study described by Boroditsky et al. (2003), which showed how the perception of objects in different languages can be affected by their gender differences. For instance, one of the words used in their study is the word “bridge,” which is masculine in Spanish and feminine in German: when asked about the descriptive properties of a bridge, Spanish speakers described bridges as being big, dangerous, long, strong, sturdy, and towering, while German speakers said they are beautiful, elegant, fragile, peaceful, pretty, and slender. While this previous research has the benefit of careful in-lab studies that explore differences in world view for one dimension (e.g., time, space) or word (e.g., bridge, sun) at a time, it also has limitations in terms of the number of experiments that can be run when subjects are being brought to the lab for every new question being asked. We aim to address this shortcoming by using the power of large-scale computational linguistics, which allows us …

Content truncated to 500 words