Cross-lingual Inference with A Chinese Entailment Graph
This work addresses cross-lingual inference for question-answering, offering incremental improvements by extending existing methods to Chinese and demonstrating complementarity between languages.
The paper tackles the problem of building Chinese entailment graphs for predicate entailment detection, introducing a novel open relation extraction method and a Chinese fine-grained entity typing dataset, and shows that an ensemble of Chinese and English graphs outperforms monolingual ones by 4.7 AUC points on a benchmark dataset.
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.