DualTKB: A Dual Learning Bridge between Text and Knowledge Base

TL;DR

This paper introduces DualTKB, a dual learning framework for unsupervised text-path and path-text transfer in Commonsense Knowledge Bases, improving KB completion and conversion to textual descriptions with weak supervision.

Contribution

It proposes a novel dual learning approach with weak supervision for bidirectional text and knowledge base transfer, enhancing KB completion and text generation.

Findings

Weak supervision significantly improves transfer quality

The method outperforms existing baselines

Proposes a new KB completion metric for generative models

Abstract

In this work, we present a dual learning approach for unsupervised text to path and path to text transfers in Commonsense Knowledge Bases (KBs). We investigate the impact of weak supervision by creating a weakly supervised dataset and show that even a slight amount of supervision can significantly improve the model performance and enable better-quality transfers. We examine different model architectures, and evaluation metrics, proposing a novel Commonsense KB completion metric tailored for generative models. Extensive experimental results show that the proposed method compares very favorably to the existing baselines. This approach is a viable step towards a more advanced system for automatic KB construction/expansion and the reverse operation of KB conversion to coherent textual descriptions.