Answering Conversational Questions on Structured Data without Logical Forms

TL;DR

This paper introduces a new method for answering conversational questions on structured data like knowledge bases and tables without relying on logical forms, using graph neural networks and pointer mechanisms.

Contribution

It proposes a novel approach that encodes structured data as graphs with Transformer-based GNNs and answers questions via pointer networks, bypassing logical form generation.

Findings

Achieved competitive results on the SQA benchmark

Effectively handles conversational references in structured data

Demonstrates the viability of graph-based encoding for QA

Abstract

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).