A Corpus-based Multidimensional Analysis of Linguistic Features of Truth and Deception

A Corpus-based Multidimensional Analysis of Linguistic Features of Truth and Deception Zheng Fanghua School of Foreign Languages, Peking University zhengfh2016@pku.edu.cn Abstract This study sets out to examine the linguis- tic difference between truthful and decep- tive texts. In order to take more linguistic features into consideration, this research applied multidimensional analysis, which can reduce many linguistic features into several factors. This study used a self-built corpus containing 100 truthful texts and 100 deceptive texts. TextMind was em- ployed to annotate these Chinese texts au- tomatically. SPSS version 20 was utilized for t-tests and multidimensional analysis. The discussion of the data was divided into two parts: word count and word per sen- tence, and multidimensional analysis. This research reveals that word count and word per sentence of deceptive discourse are significantly smaller than those of truthful discourse. The results of multidimensional analysis suggest that deceptive discourse displays a weaker performance on dimen- sions of narration, interpersonal relation- ship, and perception. Key words: Truth, Deception, Corpus, Multidimensional Analysis 1 Introduction Lying is considered as an important and frequent part of everyday social interactions of our age and some studies suggest that on average people tell one or two lies everyday (DePaulo, Kashy, Kirkendol, Wyer, & Epstein, 1996; Hancock, Thom-Santelli, & Ritchie, 2004). In Ekman’s (1992) definition, the word “deception” and “lie” can be used interchangeably, which means that a person intentionally or deliberately misleads an- other person without mentioning his purpose in advance. Ekman (1992) further explains that there are two ways to lie: to conceal and to falsify. To conceal means that the liar withholds some infor- mation that misleads the listener. To falsify means to fabricate information, to distort facts or to con- vey an opposite attitude. Since deception has become prevalent in every- day life, especially with the development of tech- nology, how to detect deception becomes a key issue. People are poor at detecting lies (Vrij, Ed- ward, Roberts & Bull, 2000). Vrij et al. mentions that when detecting lies through non-verbal cues, such as face expression, pitch of voice or speech rate, the accuracy rates usually ranges from 40% to 60%, sometimes even lower than a chance of guess (50%). In the past two decades, a growing number of studies have been done to find linguistic cues to improve the deception detection accuracy (DePau- lo, Lindsay, Malone, Muhlenbruck, Charlton, & Cooper, 2003; Newman, Pennybaker, Berry & Richards, 2003; Bond & Lee, 2005). Many linguis- tic cues have been analyzed, including pronouns, negative emotions, conjunctions, prepositions and so on. The number of these linguistic features is often calculated through Linguistic Inquiry and Word Count (LIWC), which is a computerized text analysis program (Pennebaker, Francis, & Booth, 2012). Inspired by LIWC, Computational Cyber- Psychology Lab developed TextMind, which can be used to analyze Chinese texts. These studies, however, only select a limit num- ber of linguistic features to analyze. In order to obtain a more comprehensive view, this study aims to include more linguistic features and take the multidimensional …

Content truncated to 500 words