Active learning identifies data points from a pool of unlabeled examples whose labels, if made available, are most likely to improve the predictions of a supervised model. Most research on active learning assumes that an agent has access to the entire pool of unlabeled data and can ask for labels of any data points during an initial training phase. However, when incorporated in a larger task, an agent may only be able to query some subset of the unlabeled pool. An agent can also opportunistically query for labels that may be useful in the future, even if they are not immediately relevant. In this paper, we demonstrate that this type of opportunistic active learning can improve performance in grounding natural language descriptions of everyday objects---an important skill for home and office robots. We find, with a real robot in an object identification setting, that inquisitive behavior---asking users important questions about the meanings of words that may be off-topic for the current dialog---leads to identifying the correct object more often over time.
more »
« less
Jointly Improving Parsing and Perception for Natural Language Commands through Human-Robot Dialog
In this work, we present methods for using human-robot dialog to improve language understanding for a mobile robot agent. The agent parses natural language to underlying semantic meanings and uses robotic sensors to create multi-modal models of perceptual concepts like red and heavy. The agent can be used for showing navigation routes, delivering objects to people, and relocating objects from one location to another. We use dialog clari_cation questions both to understand commands and to generate additional parsing training data. The agent employs opportunistic active learning to select questions about how words relate to objects, improving its understanding of perceptual concepts. We evaluated this agent on Amazon Mechanical Turk. After training on data induced from conversations, the agent reduced the number of dialog questions it asked while receiving higher usability ratings. Additionally, we demonstrated the agent on a robotic platform, where it learned new perceptual concepts on the y while completing a real-world task.
more »
« less
- NSF-PAR ID:
- 10160084
- Date Published:
- Journal Name:
- Journal of Artificial Intelligence Research
- Volume:
- 67
- ISSN:
- 1076-9757
- Page Range / eLocation ID:
- 327 to 374
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Natural language understanding for robotics can require substantial domain- and platform-specific engineering. For example, for mobile robots to pick-and-place objects in an environment to satisfy human commands, we can specify the language humans use to issue such commands, and connect concept words like red can to physical object properties. One way to alleviate this engineering for a new domain is to enable robots in human environments to adapt dynamically -- continually learning new language constructions and perceptual concepts. In this work, we present an end-to-end pipeline for translating natural language commands to discrete robot actions, and use clarification dialogs to jointly improve language parsing and concept grounding. We train and evaluate this agent in a virtual setting on Amazon Mechanical Turk, and we transfer the learned agent to a physical robot platform to demonstrate it in the real world.more » « less
-
Faust, Aleksandra ; Hsu, David ; Neumann, Gerhard (Ed.)Enabling human operators to interact with robotic agents using natural language would allow non-experts to intuitively instruct these agents. Towards this goal, we propose a novel Transformer-based model which enables a user to guide a robot arm through a 3D multi-step manipulation task with natural language commands. Our system maps images and commands to masks over grasp or place locations, grounding the language directly in perceptual space. In a suite of block rearrangement tasks, we show that these masks can be combined with an existing manipulation framework without re-training, greatly improving learning efficiency. Our masking model is several orders of magnitude more sample efficient than typical Transformer models, operating with hundreds, not millions, of examples. Our modular design allows us to leverage supervised and reinforcement learning, providing an easy interface for experimentation with different architectures. Our model completes block manipulation tasks with synthetic commands more often than a UNet-based baseline, and learns to localize actions correctly while creating a mapping of symbols to perceptual input that supports compositional reasoning. We provide a valuable resource for 3D manipulation instruction following research by porting an existing 3D block dataset with crowdsourced language to a simulated environment. Our method’s absolute improvement in identifying the correct block on the ported dataset demonstrates its ability to handle syntactic and lexical variation.more » « less
-
more » « less
We describe and experimentally validate a question-asking framework for machine-learned linguistic knowledge about human emotions. Using the Socratic method as a theoretical inspiration, we develop an experimental method and computational model for computers to learn subjective information about emotions by playing emotion twenty questions (EMO20Q), a game of twenty questions limited to words denoting emotions. Using human–human EMO20Q data we bootstrap a sequential Bayesian model that drives a generalized pushdown automaton-based dialog agent that further learns from 300 human–computer dialogs collected on Amazon Mechanical Turk. The human–human EMO20Q dialogs show the capability of humans to use a large, rich, subjective vocabulary of emotion words. Training on successive batches of human–computer EMO20Q dialogs shows that the automated agent is able to learn from subsequent human–computer interactions. Our results show that the training procedure enables the agent to learn a large set of emotion words. The fully trained agent successfully completes EMO20Q at 67% of human performance and 30% better than the bootstrapped agent. Even when the agent fails to guess the human opponent’s emotion word in the EMO20Q game, the agent’s behavior of searching for knowledge makes it appear human-like, which enables the agent to maintain user engagement and learn new, out-of-vocabulary words. These results lead us to conclude that the question-asking methodology and its implementation as a sequential Bayes pushdown automaton are a successful model for the cognitive abilities involved in learning, retrieving, and using emotion words by an automated agent in a dialog setting.
-
A Natural Language Interface (NLI) enables the use of human languages to interact with computer systems, including smart phones and robots. Compared to other types of interfaces, such as command line interfaces (CLIs) or graphical user interfaces (GUIs), NLIs stand to enable more people to have access to functionality behind databases or APIs as they only require knowledge of natural languages. Many NLI applications involve structured data for the domain (e.g., applications such as hotel booking, product search, and factual question answering.) Thus, to fully process user questions, in addition to natural language comprehension, understanding of structured data is also crucial for the model. In this paper, we study neural network methods for building Natural Language Interfaces (NLIs) with a focus on learning structure data representations that can generalize to novel data sources and schemata not seen at training time. Specifically, we review two tasks related to natural language interfaces: i) semantic parsing where we focus on text-to-SQL for database access, and ii) task-oriented dialog systems for API access. We survey representative methods for text-to-SQL and task-oriented dialog tasks, focusing on representing and incorporating structured data. Lastly, we present two of our original studies on structured data representation methods for NLIs to enable access to i) databases, and ii) visualization APIs.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1613/jair.1.11485
