FrameNet Annotations Alignment using Attention-based Machine Translation

摘要

This paper presents an approach to project FrameNet annotations into other languages using attention-based neural machine translation (NMT) models. The idea is to use a NMT encoder-decoder attention matrix to propose a word-to-word correspondence between the source and the target languages. We combine this word alignment along with a set of simple rules to securely project the FrameNet annotations into the target language. We successfully implemented, evaluated and analyzed this technique on the English-to-French configuration. First, we analyze the obtained corpus quantitatively and qualitatively. Then, we use existing FrameNet corpora to assert the quality of the translation. Finally, we trained a BERT-based FrameNet parser using the projected annotations and compared it to a BERT baseline. Results show modest performance gains in the French language, giving evidence to support that our approach could help to propagate FrameNet data-set on other languages. Moreover, this label projection approach can be extended to other sequence tagging tasks with minor modifications.