Data Recombination for Neural Semantic Parsing

Abstract

Modeling crisp logical regularities is crucial in semantic parsing, making it difficult for neural models with no task-specific prior knowledge to achieve good results. In this paper, we introduce data recombination, a novel framework for injecting such prior knowledge into a model. From the training data, we induce a high-precision synchronous context-free grammar, which captures important conditional independence properties commonly found in semantic parsing. We then train a sequence-to-sequence recurrent network (RNN) model with a novel attention-based copying mechanism on datapoints sampled from this grammar, thereby teaching the model about these structural properties. Data recombination improves the accuracy of our RNN model on three semantic parsing datasets, leading to new state-of-the-art performance on the standard GeoQuery dataset for models with comparable supervision.

ACL 2016

References in corpus (12)

• Sequence to Sequence Learning with Neural Networks
• Improving neural networks by preventing co-adaptation of feature detectors
• Learning to Map Sentences to Logical Form: Structured Classification with Probabilistic Categorial Grammars
• Effective Approaches to Attention-based Neural Machine Translation
• Transition-Based Dependency Parsing with Stack Long Short-Term Memory
• Language to Logical Form with Neural Attention
• Listen, Attend, and Walk: Neural Mapping of Navigational Instructions to Action Sequences
• Traversing Knowledge Graphs in Vector Space
• Learning Dependency-Based Compositional Semantics
• Pointing the Unknown Words
• Cornell SPF: Cornell Semantic Parsing Framework
• A Deep Architecture for Semantic Parsing

Cited by in corpus (14)

• A Survey of Deep Learning for Scientific Discovery
• Ape210K: A Large-Scale and Template-Rich Dataset of Math Word Problems
• Improving Grammatical Error Correction via Pre-Training a Copy-Augmented Architecture with Unlabeled Data
• CraftAssist: A Framework for Dialogue-enabled Interactive Agents
• Transfer Learning for Neural Semantic Parsing
• Why Build an Assistant in Minecraft?
• Two-Stage Synthesis Networks for Transfer Learning in Machine Comprehension
• CraftAssist Instruction Parsing: Semantic Parsing for a Minecraft Assistant
• Exploring Neural Models for Parsing Natural Language into First-Order Logic
• Teaching Machines to Converse
• Query Understanding for Natural Language Enterprise Search
• Compositional pre-training for neural semantic parsing
• Learning to Learn Semantic Parsers from Natural Language Supervision
• Few-Shot Semantic Parsing for New Predicates