Reddit Entity Linking Dataset

TL;DR

This paper presents a new Reddit-based entity linking dataset with annotated entities, analyzes annotation disagreements, and evaluates existing models, revealing their poor performance on social media data and highlighting the need for specialized models.

Contribution

The paper introduces a publicly available Reddit entity linking dataset with detailed annotations and error analysis, and evaluates existing models showing their limitations on social media text.

Findings

Existing models perform poorly on social media data.

Most errors are due to mention detection issues.

Better models are needed for social media entity linking.

Abstract

We introduce and make publicly available an entity linking dataset from Reddit that contains 17,316 linked entities, each annotated by three human annotators and then grouped into Gold, Silver, and Bronze to indicate inter-annotator agreement. We analyze the different errors and disagreements made by annotators and suggest three types of corrections to the raw data. Finally, we tested existing entity linking models that are trained and tuned on text from non-social media datasets. We find that, although these existing entity linking models perform very well on their original datasets, they perform poorly on this social media dataset. We also show that the majority of these errors can be attributed to poor performance on the mention detection subtask. These results indicate the need for better entity linking models that can be applied to the enormous amount of social media text.