A Cross-lingual Annotation Projection-based Self-supervision Approach for Open Information Extraction

Open information extraction (IE) is a weakly supervised IE paradigm that aims to extract relation-independent information from large-scale natural language documents without significant annotation efforts. A key challenge for Open IE is to achieve self-supervision, in which the training examples are automatically obtained. Although the feasibility of Open IE systems has been demonstrated for English, utilizing such techniques to build the systems for other languages is problematic because previous self-supervision approaches require language-specific knowledge. To improve the cross-language portability of Open IE systems, this paper presents a self-supervision approach that exploits parallel corpora to obtain training examples for the target language by projecting the annotations onto the source language. The merit of our method is demonstrated using a Korean Open IE system developed without any language-specific knowledge.