Compositional Data and Task Augmentation for Instruction Following

Executing natural language instructions in a physically grounded domain requires a model that understands both spatial concepts such as left of'' and above'', and the compositional language used to identify landmarks and articulate instructions relative to them. In this paper, we study instruction understanding in the blocks world domain. Given an initial arrangement of blocks and a natural language instruction, the system executes the instruction by manipulating selected blocks. The highly compositional instructions are composed of atomic components and understanding these components is a necessary step to executing the instruction. We show that while end-to-end training (supervised only by the correct block location) fails to address the challenges of this task and performs poorly on instructions involving a single atomic component, knowledge-free auxiliary signals can be used to significantly improve performance by providing supervision for the instruction's components. Specifically, we generate signals that aim at helping the model gradually understand components of the compositional instructions, as well as those that help it better understand spatial concepts, and show their benefit to the overall task for two datasets and two state-of-the-art (SOTA) models, especially when the training data is limited---which is usual in such tasks.

Generalization in Instruction Following Systems

Understanding and executing natural language instructions in a grounded domain is one of the hallmarks of artificial intelligence. In this paper, we focus on instruction understanding in the blocks world domain and investigate the language understanding abilities of two top-performing systems for the task. We aim to understand if the test performance of these models indicates an understanding of the spatial domain and of the natural language instructions relative to it, or whether they merely over-fit spurious signals in the dataset. We formulate a set of expectations one might have from an instruction following model and concretely characterize the different dimensions of robustness such a model should possess. Despite decent test performance, we find that state-of-the-art models fall short of these expectations and are extremely brittle. We then propose a learning strategy that involves data augmentation and show through extensive experiments that the proposed learning strategy yields models that are competitive on the original test set while satisfying our expectations much better.

Extraction of Common-Sense Relations from Procedural Task Instructions using BERT

Manipulation-relevant common-sense knowledge is crucial to support action-planning for complex tasks. In particular, instrumentality information of what can be done with certain tools can be used to limit the search space which is growing exponentially with the number of viable options. Typical sources for such knowledge, structured common-sense knowledge bases such as ConceptNet or WebChild, provide a limited amount of information which also varies drastically across different domains. Considering the recent success of pre-trained language models such as BERT, we investigate whether common-sense information can directly be extracted from semi-structured text with an acceptable annotation effort. Concretely, we compare the common-sense relations obtained from ConceptNet versus those extracted with BERT from large recipe databases. In this context, we propose a scoring function, based on the WordNet taxonomy to match specific terms to more general ones, enabling a rich evaluation against a set of ground-truth relations.