Enhancing Human–Computer Interaction With Cultural Nuance: A Deep Reinforcement Learning Perspective

Abstract

Human-Computer Interaction (HCI) has drawn a lot of attention lately as one of the most important and thought-provoking uses of artificial intelligence. The computer must consider a number of factors and circumstances to guarantee successful human-computer interaction. In order to improve the HCI model and create a more enduring relationship, it is essential to carefully assess each person’s temperament. The efficacy of sentiment analysis methods integrated into HCI models is examined in this paper in relation to cultural aspects. To achieve this goal, a deep reinforcement learning architecture is presented to correctly categorize people’s temperaments while taking their cultural background into account. In this paradigm, the task of replicating the affective states of basic cultures falls to each deep neural network. A wide variety of ethnic and cultural origins can be simulated by using weighted mixtures of these basic cultural models. Therefore, to accurately evaluate the temperaments of particular individuals, the proposed method uses an autonomous weighting mechanism based on learning to determine the weighted combinations of basic models. A dataset of audiovisual content related to various civilizations was used to evaluate the effectiveness of the suggested methodology. The results of the studies show that adding cultural backgrounds to temperament recognition can result in significant gains, as demonstrated by a 4.6% reduction in detection error over previous models. Furthermore, mood detection error can be decreased by up to 2.95 percent by using reinforcement learning techniques that make use of weighted combinations of basic models. With the mean absolute error of 0.129, the obtained results demonstrate that the proposed model can identify emotional states across different cultures. This demonstrates how well the model enhances the functionality of HCI models.