Abstract
Relation extraction (RE) seeks to detect and classify semantic relationshipsbetween entities, which provides useful information for many NLP applications.Since the state-of-the-art RE models require large amounts of manuallyannotated data and language-specific resources to achieve high accuracy, it isvery challenging to transfer an RE model of a resource-rich language to aresource-poor language. In this paper, we propose a new approach forcross-lingual RE model transfer based on bilingual word embedding mapping. Itprojects word embeddings from a target language to a source language, so that awell-trained source-language neural network RE model can be directly applied tothe target language. Experiment results show that the proposed approachachieves very good performance for a number of target languages on bothin-house and open datasets, using a small bilingual dictionary with only 1Kword pairs.