Multi-View Graph Representation Learning for Answering Hybrid Numerical Reasoning Question

TL;DR

This paper introduces a Multi-View Graph Encoder for hybrid question answering over financial reports, effectively capturing the relationships between text and tables to improve numerical reasoning accuracy.

Contribution

It proposes a novel Multi-View Graph Encoder that preserves the granularity and spatial structure of hybrid data, outperforming existing models on TAT-QA benchmark.

Findings

Outperforms state-of-the-art on TAT-QA benchmark

Effectively captures relations among table, text, and numerical data

Preserves original data characteristics for better reasoning

Abstract

Hybrid question answering (HybridQA) over the financial report contains both textual and tabular data, and requires the model to select the appropriate evidence for the numerical reasoning task. Existing methods based on encoder-decoder framework employ a expression tree-based decoder to solve numerical reasoning problems. However, encoders rely more on Machine Reading Comprehension (MRC) methods, which take table serialization and text splicing as input, damaging the granularity relationship between table and text as well as the spatial structure information of table itself. In order to solve these problems, the paper proposes a Multi-View Graph (MVG) Encoder to take the relations among the granularity into account and capture the relations from multiple view. By utilizing MVGE as a module, we constuct Tabular View, Relation View and Numerical View which aim to retain the original characteristics of the hybrid data. We validate our model on the publicly available table-text hybrid QA benchmark (TAT-QA) and outperform the state-of-the-art model.