Graph-to-Tree Neural Networks for Learning Structured Input-Output Translation with Applications to Semantic Parsing and Math Word Problem
This addresses the need for better handling of structural information in translation tasks like semantic parsing and math word problems, though it is incremental as it builds on existing Seq2Seq variants.
The paper tackles the problem of structured input-output translation by proposing Graph-to-Tree Neural Networks (Graph2Tree), which encode graph-structured inputs and decode tree-structured outputs, and demonstrates that it outperforms or matches state-of-the-art models on semantic parsing and math word problem tasks.
The celebrated Seq2Seq technique and its numerous variants achieve excellent performance on many tasks such as neural machine translation, semantic parsing, and math word problem solving. However, these models either only consider input objects as sequences while ignoring the important structural information for encoding, or they simply treat output objects as sequence outputs instead of structural objects for decoding. In this paper, we present a novel Graph-to-Tree Neural Networks, namely Graph2Tree consisting of a graph encoder and a hierarchical tree decoder, that encodes an augmented graph-structured input and decodes a tree-structured output. In particular, we investigated our model for solving two problems, neural semantic parsing and math word problem. Our extensive experiments demonstrate that our Graph2Tree model outperforms or matches the performance of other state-of-the-art models on these tasks.