Answering Conversational Questions on Structured Data without Logical Forms

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 5902–5910, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 5902 Answering Conversational Questions on Structured Data without Logical Forms Thomas M¨uller, Francesco Piccinno, Massimo Nicosia, Peter Shaw, Yasemin Altun Google Research {thomasmueller,piccinno,massimon,petershaw,altun}@google.com Abstract We present a novel approach to answering se- quential questions based on structured objects such as knowledge bases or tables without us- ing a logical form as an intermediate repre- sentation. We encode tables as graphs us- ing a graph neural network model based on the Transformer architecture. The answers are then selected from the encoded graph using a pointer network. This model is appropri- ate for processing conversations around struc- tured data, where the attention mechanism that selects the answers to a question can also be used to resolve conversational references. We demonstrate the validity of this approach with competitive results on the Sequential Question Answering (SQA) task (Iyyer et al., 2017). 1 Introduction In recent years, there has been significant progress on conversational question answering (QA), where questions can be meaningfully an- swered only within the context of a conversation (Iyyer et al., 2017; Choi et al., 2018; Saha et al., 2018). This line of work, as in single QA setup, falls into two main categories, (i) the answers are extracted from some text in a reading comprehen- sion setting, (ii) the answers are extracted from structured objects, such as knowledge bases or ta- bles. The latter is commonly posed as a semantic parsing task, where the goal is to map questions to some logical form which is then executed over the knowledge base to extract the answers. In semantic parsing, there is extensive work on using deep neural networks for training models over manually created logical forms in a super- vised learning setup (Jia and Liang, 2016; Ling et al., 2016; Dong and Lapata, 2018). However, creating labeled data for this task can be expen- sive and time-consuming. This problem resulted in research that investigates semantic parsing with weak supervision where training data consists of questions and answers along with the structured resources to recover the logical form representa- tions that would yield the correct answer (Liang et al., 2016; Iyyer et al., 2017). In this paper, we follow this line of research and investigate answering sequential questions with respect to structured objects. In contrast to pre- vious approaches, instead of learning the interme- diate logical forms, we propose a novel approach that encodes the structured resources, i.e. tables, along with the questions and answers from the context of the conversation. This approach allows us to handle conversational context without the definition of detailed operations or a vocabulary dependent on the logical form formalism that are required in the weakly supervised semantic pars- ing approaches. We present empirical performance of the pro- posed approach on the Sequential Question An- swering task (SQA) (Iyyer et al., 2017) which improves …

Content truncated to 500 words