ToD4IR: A Humanised Task-Oriented Dialogue System for Industrial Robots

Despite the fact that task-oriented conversation systems have received much attention from dialogue research community, only a handful of them been studied in real-world manufacturing context using industrial robots. One stumbling block is lack domain-specific discourse corpus for training these systems. Another difficulty earlier attempts to integrate natural language interfaces (such as chatbots) into sector primarily focused on task completion rates. When designing system social robots, user experience prioritized above We provide Industrial Robots Domain Wizard-of-Oz dataset (IRWoZ) overcome challenges, fully-labeled covering four robotics domains. It delivers simulated discussions between shop floor workers and with over 401 dialogues, promote language-assisted Human-Robot Interaction (HRI) settings. Small talk concepts human-to-human strategies are provided support humanlike answer generation give more adaptable environment increase engagement. Finally, we propose evaluate an end-to-end Task-oriented Dialogue (ToD4IR) two types pre-trained backbone models: GPT-2 GPT-Neo, IRWoZ dataset. ToD4IR’s performance real was validated through series trials. Our experiments demonstrate ToD4IR outperforms three downstream tasks, i.e., state tracking, act generation, response source code accessible at https://github.com/lcroy/ToD4IR reproducible research.