Exploring Demographic Language Variations to Improve Multilingual Sentiment Analysis in Social Media

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1815–1827, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Exploring Demographic Language Variations to Improve Multilingual Sentiment Analysis in Social Media Svitlana Volkova Center for Language and Speech Processing Johns Hopkins University Baltimore, MD svitlana@jhu.edu Theresa Wilson Human Language Technology Center of Excellence Johns Hopkins University Baltimore, MD taw@jhu.edu David Yarowsky Center for Language and Speech Processing Johns Hopkins University Baltimore, MD yarowsky@cs.jhu.edu Abstract Different demographics, e.g., gender or age, can demonstrate substantial variation in their language use, particularly in informal contexts such as social media. In this paper we focus on learning gender differences in the use of sub- jective language in English, Spanish, and Rus- sian Twitter data, and explore cross-cultural differences in emoticon and hashtag use for male and female users. We show that gen- der differences in subjective language can ef- fectively be used to improve sentiment anal- ysis, and in particular, polarity classification for Spanish and Russian. Our results show statistically significant relative F-measure im- provement over the gender-independent base- line 1.5% and 1% for Russian, 2% and 0.5% for Spanish, and 2.5% and 5% for English for polarity and subjectivity classification. 1 Introduction Sociolinguistics and dialectology have been study- ing the relationships between language and speech at the phonological, lexical and morphosyntactic lev- els and social identity for decades (Picard, 1997; Gefen and Ridings, 2005; Holmes and Meyerhoff, 2004; Macaulay, 2006; Tagliamonte, 2006). Re- cent studies have focused on exploring demographic language variations in personal email communica- tion, blog posts, and public discussions (Boneva et al., 2001; Mohammad and Yang, 2011; Eisenstein et al., 2010; O’Connor et al., 2010; Bamman et al., 2012). However, one area that remains largely unex- plored is the effect of demographic language varia- tion on subjective language use, and whether these differences may be exploited for automatic senti- ment analysis. With the growing commercial im- portance of applications such as personalized rec- ommender systems and targeted advertising (Fan and Chang, 2009), detecting helpful product review (Ott et al., 2011), tracking sentiment in real time (Resnik, 2013), and large-scale, low-cost, passive polling (O’Connor et al., 2010), we believe that sen- timent analysis guided by user demographics is a very important direction for research. In this paper, we focus on gender demographics and language in social media to investigate differ- ences in the language used to express opinions in Twitter for three languages: English, Spanish, and Russian. We focus on Twitter data because of its vol- ume, dynamic nature, and diverse population world- wide.1 We find that some words are more or less likely to be positive or negative in context depend- ing on the the gender of the author. For example, the word weakness is more likely to be used in a pos- itive way by women (Chocolate is my weakness!) but in a negative way by men (Clearly they know our weakness. Argggg). The Russian word достичь (achieve) is used in a positive …

Content truncated to 500 words