Finding Zelig in Text: A Measure for Normalising Linguistic Accommodation

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 455–465, Dublin, Ireland, August 23-29 2014. Finding Zelig in Text: A Measure for Normalising Linguistic Accommodation Simon Jones1, Rachel Cotterill2, Nigel Dewdney2, Kate Muir3, and Adam Joinson3 1Department of Computer Science, University of Bath, BA2 7AY. s.jones2@bath.ac.uk 2Department of Computer Science, University of Sheffield, S1 4DP. r.cotterill@sheffield.ac.uk, acp08njd@sheffield.ac.uk 3Behavioural Research Lab, Faculty of Business and Law, University of the West of England, Bristol, BS16 1QY. kate.muir@uwe.ac.uk, adam.joinson@uwe.ac.uk Abstract Linguistic accommodation is a recognised indicator of social power and social distance. How- ever, different individuals will vary their language to different degrees, and only a portion of this variance will be due to accommodation. This paper presents the Zelig Quotient, a method of normalising linguistic variation towards a particular individual, using an author’s other com- munications as a baseline, thence to derive a method for identifying accommodation-induced variation with statistical significance. This work provides a platform for future efforts towards examining the importance of such phenomena in large communications datasets. 1 Introduction “Zelig...protects himself by becoming like whoever he is around.” - The Narrator, Zelig (Allen, 1983) When people converse, they often become more alike in their language in many different dimensions (Garrod and Pickering, 2004). This can include similarity in pronunciation (Giles, 1973), speech rates (Street, 1984), pause and utterance duration (Cappella, 1979), and volume (Natale, 1975). Similarly, in written communications people often converge in terms of features such as linguistic style (Danescu- Niculescu-Mizil and Lee, 2011), vocabulary, and syntax (Scissors et al., 2008). Communication Ac- commodation Theory (Giles and Ogay, 2007) proposes that interactants can adjust their communication style, such as accent, vocabulary, and use of jargon to sound more (convergence) or less (divergence) like the other person. Individuals typically converge to signal affinity with their interlocutor, and diverge to show interpersonal or social distance. One area which has been largely overlooked, to date, is the role played by an individual’s inherent tendency to accommodate (or not). We propose that some people are more apt than others to change their typical linguistic style to converge to that of their conversational partner. This paper introduces the Zelig Quotient, which is a new method for capturing the degree to which the variation in an individual’s language use can be explained by their accommodation towards the style of their interlocutor. Using this score, it is then possible to measure the significance of an individual’s accommodation within a specific communication pair: does each individual accommodate more or less than their personal norm? In this paper, we firstly consider existing computational measures of linguistic accommodation. Al- though useful in measuring accommodation of specific linguistic features within dialog, current measures This work is licensed under a Creative Commons Attribution 4.0 International Licence. Page numbers and proceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ 455 do not permit examination of the role of an individual’s inherent (or latent) predisposition to accommo- date their linguistic style. Further, another area that has …

Content truncated to 500 words