We present a novel approach to answering sequential questions based on structured objects such as knowledge bases or tables without using a logical form as an intermediate representation. We encode tables as graphs using 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 appropriate for processing conversations around structured 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).
@article{arxiv.1908.11787,
title = {Answering Conversational Questions on Structured Data without Logical Forms},
author = {Thomas Müller and Francesco Piccinno and Massimo Nicosia and Peter Shaw and Yasemin Altun},
journal= {arXiv preprint arXiv:1908.11787},
year = {2019}
}