Editing-Based SQL Query Generation for Cross-Domain Context-Dependent Questions

We focus on the cross-domain contextdependent text-to-SQL generation task. Based on the observation that adjacent natural language questions are often linguistically dependent and their corresponding SQL queries tend to overlap, we utilize the interaction history by editing the previous predicted query to improve the generation quality. Our editing mechanism views SQL as sequences and reuses generation results at the token level in a simple manner. It is flexible to change individual tokens and robust to error propagation. Furthermore, to deal with complex table structures in different domains, we employ an utterance-table encoder and a table-aware decoder to incorporate the context of the user utterance and the table schema. We evaluate our approach on the SParC dataset and demonstrate the benefit of editing compared with the state-of-the-art baselines which generate SQL from scratch. Our code is available at https://github.com/ ryanzhumich/sparc_atis_pytorch.

Introduction. Generating SQL queries from user utterances is an important task to help end users acquire information from databases. In a real-world application, users often access information in a multi-turn interaction with the system by asking a sequence of related questions. As the interaction proceeds, the user often makes reference to the relevant mentions in the history or omits previously conveyed information assuming it is known to the system. Therefore, in the context-dependent scenario, the contextual history is crucial to understand the follow-up questions from users, and the system often needs to reproduce partial sequences generated in previous turns. Recently, Suhr et al. (2018) proposes a context-dependent text-to-SQL model including an interaction-level encoder and an attention mechanism over previous utterances. To reuse what has been generated, they propose to copy complete segments from the previous query.