A multi-task semi-supervised framework for Text2Graph & Graph2Text

TL;DR

This paper introduces a multi-task semi-supervised T5-based framework that jointly learns to extract graphs from text and generate text from graphs, improving performance and domain adaptability in Text2Graph and Graph2Text tasks.

Contribution

It presents a novel cycle training approach with non-parallel data, enhancing cross-domain consistency and surpassing state-of-the-art results in Text2Graph and Graph2Text tasks.

Findings

Outperforms unsupervised state-of-the-art in WebNLG tasks

More consistent across seen and unseen domains

Easily adaptable to new domains with non-parallel data

Abstract

The Artificial Intelligence industry regularly develops applications that mostly rely on Knowledge Bases, a data repository about specific, or general, domains, usually represented in a graph shape. Similar to other databases, they face two main challenges: information ingestion and information retrieval. We approach these challenges by jointly learning graph extraction from text and text generation from graphs. The proposed solution, a T5 architecture, is trained in a multi-task semi-supervised environment, with our collected non-parallel data, following a cycle training regime. Experiments on WebNLG dataset show that our approach surpasses unsupervised state-of-the-art results in text-to-graph and graph-to-text. More relevantly, our framework is more consistent across seen and unseen domains than supervised models. The resulting model can be easily trained in any new domain with non-parallel data, by simply adding text and graphs about it, in our cycle framework.