My caption 😄

Neural Open Information Extraction with Transformers


In this work, we model the Open Information Extraction problem as a sequence to sequence transduction task. We use a learning based approach to train a transformer encoder-decoder architecture to extract relational triples using a large training set bootstrapped from a rule based extractor. We show that our Open IE system significantly outperforms several existing Open IE tools on a large benchmark dataset and is competitive with the state of the art, without the dependencies on other NLP tools.

Technical Report