Neural Program Search: Solving Programming Tasks from Description and Examples
This addresses program synthesis from natural language, which could aid non-programmers, but appears incremental as it builds on existing deep learning and synthesis methods.
The authors tackled the problem of generating programs from natural language descriptions and input/output examples by developing Neural Program Search, which combines deep learning and program synthesis with a domain-specific language and Seq2Tree-guided search. They showed it significantly outperforms a sequence-to-sequence baseline on a semi-synthetic dataset.
We present a Neural Program Search, an algorithm to generate programs from natural language description and a small number of input/output examples. The algorithm combines methods from Deep Learning and Program Synthesis fields by designing rich domain-specific language (DSL) and defining efficient search algorithm guided by a Seq2Tree model on it. To evaluate the quality of the approach we also present a semi-synthetic dataset of descriptions with test examples and corresponding programs. We show that our algorithm significantly outperforms a sequence-to-sequence model with attention baseline.