Execution-Guided Neural Program Decoding

We present a neural semantic parser that translates natural language questions into executable SQL queries with two key ideas. First, we develop an encoder-decoder model, where the decoder uses a simple type system of SQL to constraint the output prediction, and propose a value-based loss when copying from input tokens. Second, we explore using the execution semantics of SQL to re-pair decoded programs that result in runtime error return empty result. We propose two model-agnostics repair approaches, an ensemble model and a local program repair, and demonstrate their effectiveness over the original model. We evaluate our model on the WikiSQL dataset and show that our model achieves close to state-of-the-art results with lesser model complexity.