Syntactic Multi-view Learning for Open Information Extraction

TL;DR

This paper introduces a neural OpenIE approach that leverages both constituency and dependency syntactic structures through multi-view learning, improving relational tuple extraction from sentences.

Contribution

It models constituency and dependency trees as word-level graphs and fuses them with semantic representations using multi-view learning, a novel integration for OpenIE.

Findings

Both syntactic structures improve extraction performance.

Multi-view learning effectively combines heterogeneous syntactic information.

The approach outperforms previous models on benchmark datasets.

Abstract

Open Information Extraction (OpenIE) aims to extract relational tuples from open-domain sentences. Traditional rule-based or statistical models have been developed based on syntactic structures of sentences, identified by syntactic parsers. However, previous neural OpenIE models under-explore the useful syntactic information. In this paper, we model both constituency and dependency trees into word-level graphs, and enable neural OpenIE to learn from the syntactic structures. To better fuse heterogeneous information from both graphs, we adopt multi-view learning to capture multiple relationships from them. Finally, the finetuned constituency and dependency representations are aggregated with sentential semantic representations for tuple generation. Experiments show that both constituency and dependency information, and the multi-view learning are effective.