Open-Vocabulary Semantic Parsing with both Distributional Statistics and Formal Knowledge

Semantic parsers map language onto executable statements in a fixed schema. This mapping allows them to effectively leverage the information contained in large, formal knowledge bases (e.g., Freebase) to answer questions, but it is also fundamentally limiting---semantic parsers can only represent language that falls within their manually produced schema. Recently proposed methods for open vocabulary semantic parsing overcome this limitation by learning execution models for arbitrary language. However, all prior approaches to open vocabulary semantic parsing are purely distributional, making no use of any underlying knowledge base. We show how to combine the benefits of both of these approaches by incorporating knowledge base information into open vocabulary semantic parsing models, improving mean average precision on an open-domain natural language query task by more than 120%.