Cross-lingual Inference with A Chinese Entailment Graph

TL;DR

This paper introduces a novel pipeline for constructing Chinese entailment graphs, leveraging a new open relation extraction method and a Chinese entity typing dataset, demonstrating cross-lingual benefits in entailment detection.

Contribution

It presents the first Chinese entailment graph pipeline, including a high-recall open relation extraction method and a Chinese entity typing dataset, improving cross-lingual inference performance.

Findings

Chinese entailment graph outperforms monolingual graphs

Ensemble of Chinese and English graphs improves results

Achieves 4.7 AUC points higher than previous state-of-the-art

Abstract

Predicate entailment detection is a crucial task for question-answering from text, where previous work has explored unsupervised learning of entailment graphs from typed open relation triples. In this paper, we present the first pipeline for building Chinese entailment graphs, which involves a novel high-recall open relation extraction (ORE) method and the first Chinese fine-grained entity typing dataset under the FIGER type ontology. Through experiments on the Levy-Holt dataset, we verify the strength of our Chinese entailment graph, and reveal the cross-lingual complementarity: on the parallel Levy-Holt dataset, an ensemble of Chinese and English entailment graphs outperforms both monolingual graphs, and raises unsupervised SOTA by 4.7 AUC points.