Memory-Based Semantic Parsing

Abstract We present a memory-based model for context- dependent semantic parsing. Previous approaches focus on enabling the decoder to copy or modify parse from previous utterance, assuming there is dependency between current and parses. In this work, we propose represent contextual information using an external memory. learn context memory controller that manages by maintaining cumulative meaning of sequential user utterances. evaluate our approach three parsing benchmarks. Experimental results show can better process context-dependent demonstrates improved performance without task-specific decoders.