Neural-network-based Memory for a Social Robot

Abstract

Many recent studies have shown that behaviors and interaction logic for social robots can be learned automatically from natural examples of human-human interaction by machine learning algorithms, with minimal input from human designers [1-4]. In this work, we exceed the capabilities of the previous approaches by giving the robot memory. In earlier work, the robot's actions were decided based only on a narrow temporal window of the current interaction context. However, human behaviors often depend on more temporally distant events in the interaction history. Thus, we raise the question of whether (and how) an automated behavior learning system can learn a memory representation of interaction history within a simulated camera shop scenario. An analysis of the types of memory-setting and memory-dependent actions that occur in the camera shop scenario is presented. Then, to create more examples of such actions for evaluating a shopkeeper robot behavior learning system, an interaction dataset is simulated. A Gated Recurrent Unit (GRU) neural network architecture is applied in the behavior learning system, which learns a memory representation for performing memory-dependent actions. In an offline evaluation, the GRU system significantly outperformed a without-memory baseline system at generating appropriate memory-dependent actions. Finally, an analysis of the GRU architecture's memory representation is presented. CCS Concepts: • Human-centered computing → Human computer interaction (HCI); • Computing methodologies → Learning from demonstrations; Neural networks;

Landscape Preference of Rural Cultural Tourism Tourists Based on GIS Spatial Statistical Technology and Emotion Analysis

“…To sum up, it can be seen that the current landscape preference analysis system of rural cultural tourism tourists with specific landscape preference as the database generally has the problems of poor recognition effect, low recognition accuracy, poor stability, and low data utilization rate in practical application [14][15][16]. On the other hand, in the existing landscape preference analysis system of rural cultural tourism tourists, the vast majority of identification methods can only identify a single rural cultural tourism and cannot identify rural cultural tourism with obvious differences, so they do not have the characteristics of intelligence [17][18][19]. In addition, the utilization rate and data mining effect of the obtained landscape preference data of rural cultural tourism tourists in the process of identification are also very poor [18].…”

Systematic review of training methods for conversational systems: the potential of datasets validated with user experience

“…An interesting method in the context of learning dialogue policies is applied by Doering, Kanda, & Ishiguro (2019). In their work, a physical scenario is configured to collect human-human interactions, which are then used to simulate speech interactions in purchasing scenarios.…”

Data-Driven Imitation Learning for a Shopkeeper Robot with Periodically Changing Product Information

“…Doering et al [10] focused on the problem of modeling the hidden structure of interaction, which enabled resolution of ambiguous customer speech. Doering et al [11] showed how gated recurrent neural networks could be used to learn a memory model of customer behavior. In addition, Doering et al [12] presented a system that could explore different robot behaviors and learn online, which resulted in more varied, interesting robot behaviors than previous approaches and enabled some customization to customers' individual differences.…”