Detecting Concealed Information in Text and Speech

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 402–412 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 402 Detecting Concealed Information in Text and Speech Shengli Hu Cornell University / Ithaca, New York 14850 sh2264@cornell.edu Abstract Motivated by infamous cheating scandals in various industries and political events, we ad- dress the problem of detecting concealed in- formation in technical settings. In this work, we explore acoustic-prosodic and linguistic in- dicators of information concealment by col- lecting a unique corpus of professionals prac- ticing for oral exams while concealing infor- mation. We reveal subtle signs of concealed information in speech and text, compare, and contrast them with those in deception detec- tion literature, thus uncovering the link be- tween concealing information and deception. We then present a series of experiments that automatically detect concealed information from text and speech. We compare the use of acoustic-prosodic, linguistic, and individ- ual feature sets, using different machine learn- ing models. Finally, we present a multi-task learning framework with acoustic, linguistic, and individual features, that outperforms hu- man performance by over 15%. 1 Introduction In 2018, a cheating scandal (Mobley, 2018) at the world’s most notoriously difficult verbal exam for wine professionals shook the global wine indus- try — answers were found leaked by some exam- iners to candidates beforehand — and all results were invalidated; in 2016, with questions leak- ing ahead of political campaigns (Wemple, 2016), CNN faced a grave scandal from which only more controversies ensued; in 2000, the notorious po- tential debate leak (Bruni and Van Natta, 2000) in- between the Bush and the Gore campaigns drew the attention of F.B.I. investigators. What all of the three scandals share in common is the fact that it had been difficult to accurately identify who and to whom leaked the critical information, because the party who unfairly obtained the information tried their best to conceal and pretend otherwise. Despite the importance and potential impact of detecting concealed information, research on de- tecting concealed information has been scarce. It is partly because large-scale datasets with ground truth labels of information concealment are diffi- cult to come by. It is only in rare cases can we verify the existence of concealed information in the wild. From the perspective of information attainment and revelation, deception and concealing informa- tion are correlated ambiguously. In Table 1, we clarify the difference between the two important concepts with an information grid. When we pos- sess the critical information but appear not in pos- session, we are concealing information; whereas in contrast, when we do not possess the informa- tion but pretend we are in the know, we are de- ceiving. Despite the proliferation of deception de- tection studies in text and speech, research on the closely related problem of detecting concealed in- formation has been sparse. The Information Grid Appearance Information No Information Truth Information Honesty Concealed Information No Information Deception Honesty Table 1: The Information Grid: Concealed Information …

Content truncated to 500 words