Objectives

The goal of this workshop is to bring together researchers from robotics, natural language processing, machine learning, and cognitive science to examine the challenges and opportunities emerging from the interdisciplinary research field covering language and robotics. This goal is motivated by two fundamental observations.

First, the future evolution of robotics requires to get closer to the natural human communication, and to demonstrate similar adaptability and flexibility in language use. Robots, likewise humans, should be able to adapt to each person they talk to, not using identical stereotypical sentences for each interaction. Moreover, humans are not only able to use language but also able to learn it. In order to have similar proficiency, robotic systems may have to learn it as well, probably not exactly in the same way, but it is rather unlikely that it would rely only on an ungrounded and disembodied language module identical to any robot. Furthermore, input data received by language learners is not written text data, but multimodal sensorimotor information including speech signal, haptic information, visual information, etc. Language learning strategies in real-world environments which are full of uncertainty would need to extract the best of multimodal information available. Making this learning and understanding of utterances possible, in a real-world environment with a situated and embodied system, is a key challenge for natural language processing.

Second, robots need to involve natural language processing and semantic understanding of environments to communicate and collaborate with people, i.e., users. Robots interacting with human users using speech signals try to behave correctly based on the users' speech commands in a real-environment, e.g., the RoboCup@Home environment. However, we still have many challenges in this field, human-robot interaction using natural language. Many applications are still using hand-written rules. Deep learning-based methods including sequence to sequence are giving new approaches to the field of human-robot interaction. It is still immature to be used in the real-world tasks. Language acquisition by robots is also still in a preliminary stage, though symbol emergence/grounding in robotics and its related research field have accelerated the research about (data-driven) language acquisition by robots and produced significant achievements, during this decade.

Recent advances in machine learning techniques, including deep learning and hierarchical Bayesian modeling, are providing us with new possibilities to integrate high-level and low-level cognitive capabilities in robotics. Such a hierarchical integration of cognitive capabilities is required to enable a robot to use language to communicate and collaborate with people in the real-world environment. Because language is highly dependent on the context, this hierarchical integration in both bottom-up and top-down processes is also needed to make language processing reliable: the first-pass “perceived” phonemes/words may be updated later based on higher level priors or context.

In this workshop, we will investigate how we can create a robot that can acquire, use and understand language. Future robots are expected to have such abilities in a wide range of human-robot interactions. To this end, we aim to share knowledge about the state-of-the-art machine learning methods that contribute to modeling language-related capabilities in robotics, and to exchange views among cutting-edge robotics researchers with a special emphasis on language in robotics. The workshop will include keynote presentations from established researchers in robotics, machine learning, natural language processing, and cognitive science. There will be a poster session highlighting contributed papers throughout the day.

We believe the topic of this workshop is timely and necessary for the IEEE-RAS community.