Naturalizing a Programming Language via Interactive Learning

By Sida I. Wang, Samuel Ginn, Percy Liang, Christopher D. Manning. 2017

80 citations recorded by the source corpus.

Our goal is to create a convenient natural language interface for performing well-specified but complex actions such as analyzing data, manipulating text, and querying databases. However, existing natural language interfaces for such tasks are quite primitive compared to the power one wields with a programming language. To bridge this gap, we start with a core programming language and allow users to “naturalize” the core language incrementally by defining alternative, more natural syntax and increasingly complex concepts in terms of compositions of simpler ones. In a voxel world, we show that a community of users can simultaneously teach a common system a diverse language and use it to build hundreds of complex voxel structures. Over the course of three days, these users went from using only the core language to using the naturalized language in 85.9% of the last 10K utterances.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

ExpBERT: Representation Engineering with Natural Language Explanations (2020, background): “In a related line of work, Wang et al. (2017), users teach a system high-level concepts via language.”
Training Classifiers with Natural Language Explanations (2018, background): “Wang et al. (2017) convert natural language definitions to rules in a semantic parser to build up progressively higher-level concepts.”
LILA: Language-Informed Latent Actions (2021, background): “…disambiguate with language, and the integration of language within the latent actions framework is a strong research contribution, and hope that future work looks to dynamic states – perhaps by leveraging visual latent actions [28] – and to adapting to new utterances and tasks dynamically [43, 52].”
Learning to Speak and Act in a Fantasy Text Adventure Game (2019, methodology): “, 2016) and a more open ended building task using a grid of voxels (Wang et al., 2017).”
Mastering the Dungeon: Grounded Language Learning by Mechanical Turker Descent (2017, neutral): “Some works that do utilize real natural language and interaction include Wang et al. (2016), where language is learnt to solve block puzzles, and Wang et al. (2017) where language is learnt to draw voxel images, which are both quite different to our case of studying text adventure games.”
An Application-Independent Approach to Building Task-Oriented Chatbots with Interactive Continual Learning (2020, background): “CML is related to interactive language learning [44, 43], but unlike CML, they do not allow learner to acquire knowledge by asking questions to users in multi-turn NL dialogues.”
ELLA: Exploration through Learned Language Abstraction (2021, background): “Simplified environments also make investigating concepts such as language abstraction possible, and these concepts can generalize to real human-AI interactions (Wang et al., 2017; Karamcheti et al., 2020).”
Learning Adaptive Language Interfaces through Decomposition (2020, methodology): “This is similar to prior work that defines a set of canonical utterances (Wang et al., 2015), or a core grammar (Wang et al., 2017).”
Towards a Continuous Knowledge Learning Engine for Chatbots (2018, background): “Our work is related to interactive language learning (ILL) (Wang et al., 2016b, 2017), but these are not about KB completion.”
Language to Action: Towards Interactive Task Learning with Physical Agents (2018, background): “Tasks are compositional in nature which can be captured by grammars [Liu et al. , 2016], Hierarchical Task Networks [Hogg et al. , 2009], or internal programming language [Wang et al. , 2017], etc.”
Why Build an Assistant in Minecraft? (2019, background): “On the other hand, there are several systems designed explicitly for self improvement, for example [63, 88].”
Lifelong and Interactive Learning of Factual Knowledge in Dialogues (2019, background): “CILK is related to interactive language learning (Wang et al., 2016, 2017), which is mainly about language grounding, not about knowledge learning.”
Continual Adaptation for Efficient Machine Communication (2019, background): “also related to broader efforts to ground language learning and emergent communication in usage, where artificial agents are trained to use language from scratch by playing interactive reference games (Wang et al., 2016; Lazaridou et al., 2016; Wang et al., 2017; Chevalier-Boisvert et al., 2019).”