Applying Ensemble Learning Techniques and Neural Networks to Deceptive and Truthful Information Detection Task in the Flow of Speech

Abstract

This paper presents the results of experiments on applying ensemble learning techniques and neural networks to a paralinguistic analysis of deceptive and truthful statements in the flow of speech. Based on an analysis and comparison of different approaches to the issue, we propose using a mixture of such methods. The Real-Life Trial Deception Detection Dataset was used for both training and testing. All the experiments were performed using 10-fold cross-validation. Using two-layer neural networks, k-nearest neighbor, random forest for evaluating and principal component analysis methods for preprocessing, results in UAR of 65.0% and 70.0%, in the case of average and majority voting correspondingly.