Unified Semantic Parsing with Weak Supervision

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4801–4810 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 4801 Unified Semantic Parsing with Weak Supervision Priyanka Agrawal*, Parag Jain, Ayushi Dalmia, Abhishek Bansal, Ashish Mittal, Karthik Sankaranarayanan IBM Research AI *pagrawal.ml@gmail.com {pajain34, adalmi08, abbansal, arakeshk, kartsank}@in.ibm.com Abstract Semantic parsing over multiple knowledge bases enables a parser to exploit structural sim- ilarities of programs across the multiple do- mains. However, the fundamental challenge lies in obtaining high-quality annotations of (utterance, program) pairs across various do- mains needed for training such models. To overcome this, we propose a novel framework to build a unified multi-domain enabled se- mantic parser trained only with weak supervi- sion (denotations). Weakly supervised training is particularly arduous as the program search space grows exponentially in a multi-domain setting. To solve this, we incorporate a multi- policy distillation mechanism in which we first train domain-specific semantic parsers (teach- ers) using weak supervision in the absence of the ground truth programs, followed by train- ing a single unified parser (student) from the domain specific policies obtained from these teachers. The resultant semantic parser is not only compact but also generalizes better, and generates more accurate programs. It fur- ther does not require the user to provide a domain label while querying. On the stan- dard OVERNIGHT dataset (containing multiple domains), we demonstrate that the proposed model improves performance by 20% in terms of denotation accuracy in comparison to base- line techniques. 1 Introduction Semantic parsing is the task of converting natural language utterances into machine executable pro- grams such as SQL, lambda logical form (Liang, 2013). This has been a classical area of re- search in natural language processing (NLP) with earlier works primarily utilizing rule based ap- proaches (Woods, 1973) or grammar based ap- proaches (Lafferty et al., 2001; Kwiatkowski et al., Figure 1: Examples for natural language utterances with linguistic variations in two different domains that share structural regularity (Source: OVERNIGHT dataset). Note that in this setup, we do not use ground truth parses for training the semantic parser. 2011; Zettlemoyer and Collins, 2005, 2007). Re- cently, there has been a surge in neural encoder- decoder techniques which are trained with in- put utterances and corresponding annotated out- put programs (Dong and Lapata, 2016; Jia and Liang, 2016). However, the performance of these strongly supervised methods is restricted by the size and the diversity of training data i.e. natural language utterances and their corresponding anno- tated logical forms. This has motivated the work on applying weak supervision based approaches (Clarke et al., 2010; Liang et al., 2017; Neelakan- tan et al., 2016; Chen et al., 2018), which use de- notations i.e. the final answers obtained upon ex- ecuting a program on the knowledge base and use REINFORCE (Williams, 1992; Norouzi et al., 2016), to guide the network to learn its semantic parsing policy (see Figure 3(a)). Another line of work (Goldman et al., 2018; Cheng and Lapata, 2018) …

Content truncated to 500 words