An Empirical Analysis of Formality in Online Communication

An Empirical Analysis of Formality in Online Communication Ellie Pavlick University of Pennsylvania∗ epavlick@seas.upenn.edu Joel Tetreault Yahoo Labs tetreaul@yahoo-inc.com Abstract This paper presents an empirical study of linguistic formality. We perform an analy- sis of humans’ perceptions of formality in four different genres. These findings are used to develop a statistical model for pre- dicting formality, which is evaluated un- der different feature settings and genres. We apply our model to an investigation of formality in online discussion forums, and present findings consistent with theories of formality and linguistic coordination. 1 Introduction Language consists of much more than just con- tent. Consider the following two sentences: 1. Those recommendations were unsolicited and undesirable. 2. that’s the stupidest suggestion EVER. Both sentences communicate the same idea, but the first is substantially more formal. Such stylistic differences often have a larger impact on how the hearer understands the sentence than the literal meaning does (Hovy, 1987). Full natural language understanding requires comprehending this stylistic aspect of meaning. To enable real advancements in dialog systems, information extraction, and human-computer interaction, computers need to understand the entirety of what humans say, both the literal and the non-literal. In this paper, we focus on the ∗ Research performed while at Yahoo Labs. particular stylistic dimension illustrated above: formality. Formality has long been of interest to linguists and sociolinguists, who have observed that it subsumes a range of dimensions of style in- cluding serious-trivial, polite-casual, and level of shared knowledge (Irvine, 1979; Brown and Fraser, 1979). The formal-informal dimension has even been called the “most important di- mension of variation between styles” (Heylighen and Dewaele, 1999). A speaker’s level of formal- ity can reveal information about their familiar- ity with a person, opinions of a topic, and goals for an interaction (Hovy, 1987; Endrass et al., 2011). As a result, the ability to recognize for- mality is an integral part of dialogue systems (Mairesse, 2008; Mairesse and Walker, 2011; Battaglino and Bickmore, 2015), sociolinguistic analyses (Danescu-Niculescu-Mizil et al., 2012; Justo et al., 2014; Krishnan and Eisenstein, 2015), human-computer interaction (Johnson et al., 2005; Khosmood and Walker, 2010), summa- rization (Sidhaye and Cheung, 2015), and au- tomatic writing assessment (Felice and Deane, 2012). Formality can also indicate context- independent, universal statements (Heylighen and Dewaele, 1999), making formality detection relevant for tasks such as knowledge base popu- lation (Suh et al., 2006; Reiter and Frank, 2010) and textual entailment (Dagan et al., 2006). This paper investigates formality in online written communication. The contributions are as follows: 1) We provide an analysis of humans’ subjective perceptions of formality in four dif- ferent genres. We highlight areas of high and low agreement and extract patterns that consis- 61 Transactions of the Association for Computational Linguistics, vol. 4, pp. 61–74, 2016. Action Editor: Janyce Wiebe and Kristina Toutanova. Submission batch: 10/2015; Revision batch: 12/2015; Published 3/2016. c⃝2016 Association for Computational Linguistics. Distributed under a CC-BY 4.0 license. tently differentiate formal from informal text. 2) We develop a state-of-the-art statistical model for …

Content truncated to 500 words