Estudio de los rasgos lingüísticos de la mentira en el medio escritoun análisis contrastivo inglés-español

Abstract

The main aim of this PhD thesis is to analyse the linguistic cues to deception in written language both in English and Spanish, performing a contrastive analysis between both languages. For this purpose, several automatic classification experiments have been performed on two ad-hoc corpora in both languages, in order to check whether the texts could be successfully classified on the basis of their truth value. In the first set of experiments, a machine learning technique has been applied on the data and compared to a Bag-of-Words model, obtaining a maximum rate of 78.5% for English and 84.5% for Spanish. The second experiment involved statistical techniques, namely discriminant function analysis and binary logistic regression, and the results obtained proved remarkably successful too. In addition, they confirm the leading role in deception detection of parameters such as text length, self-references, insight and exclusive words. Keywords: computational linguistics, deception detection, contrastive analysis, automatic classification.