An Empirical Study on Finding Spans

TL;DR

This paper empirically compares different span finding methods for information extraction, revealing that no single approach is best universally and offering insights on how task properties influence method performance.

Contribution

It provides a comprehensive empirical analysis of span finding techniques, highlighting the impact of task-specific factors on their effectiveness.

Findings

Tagging approaches yield higher precision

Span enumeration and boundary prediction offer higher recall

Span type information benefits boundary prediction

Abstract

We present an empirical study on methods for span finding, the selection of consecutive tokens in text for some downstream tasks. We focus on approaches that can be employed in training end-to-end information extraction systems, and find there is no definitive solution without considering task properties, and provide our observations to help with future design choices: 1) a tagging approach often yields higher precision while span enumeration and boundary prediction provide higher recall; 2) span type information can benefit a boundary prediction approach; 3) additional contextualization does not help span finding in most cases.