SpanIE: A Span-Based Approach to OpenIE

We introduce SpanIE, a span-based approach to solve the task of Open Information Extraction. SpanIE is a two-stage framework, providing a new way to leverage linguistic knowledge in proposition extraction from unstructured text. Our system recovers the underlying argument-predicate structure of a sentence from its dependency tree, and relies on the resulting sentence representation to solve the OpenIE task. The intermediate representation generated by SpanIE, is meant to benefit the information extraction process, allowing for accurate and expressive extractions. We conduct a preliminary comparative evaluation and present the experimental results to prove the effectiveness of the proposed approach. Using the benchmark framework for OIE, we show that our implementation achieves comparable performance with the state-of-the-art Open IE systems on the relation extraction task.